InFocus

 

The big picture: Large language models (LLMs) have vast potential to improve policing by enhancing knowledge access, decision making, and operational efficiency.

Like many other sectors of society, policing is increasingly incorporating emerging artificial intelligence (AI) technologies. OpenAI released ChatGPT in late 2022 and quickly reached 100 million users in just two months. Then, almost overnight, an easy-to-use AI tool—specifically a large language model (LLM)—became freely available to the public, sparking widespread AI adoption.

LLMs allow users to understand complex, dense information through a plain language prompt. This technology simplifies tasks that once took hours of manual processing. For example, spreadsheets can easily be analyzed, and long documents can be summarized, saving minutes or hours of reading. The significant time savings can improve efficiency and enable quicker insights from data and documents.

The integration of technology within law enforcement agencies is not a new development. For example, Computer-Aided Dispatch (CAD) databases, Record Management Systems (RMS), and body-worn cameras (BWC) are considered standard technology in the field, with CAD systems having been in use for decades. Certain existing law enforcement technologies, such as gunshot detection systems and predictive policing models, could rightfully be considered narrow applications of AI because they can create actionable knowledge from raw data.

Using LLMs in combination with natural language processing (NLP) opens even more possibilities for law enforcement agencies.

 

What we know: AI’s rapid advancement will present challenges, but that doesn’t mean it shouldn’t be used.

Modern society has a “pacing problem” where emerging technologies advance so rapidly that legal, economic, social, and personal domains are unable to adapt on the same timeline. For example, in law enforcement, BWCs combined with NLP allow officers to automate report writing to the point that most of the narrative portion of the report is completed from the BWC transcript. While the desired end is met (the report), the new process raises questions about traditions and accountability. Without the luxury of time for courts, researchers, and society to weigh in on these issues, practitioners and policymakers must decide how to implement and manage these new developments.

The flexibility of LLMs is what makes their use in policing a new frontier. Similar to many information-rich fields, policing is ripe for the widespread application and integration of AI, specifically LLMs. Fueling the information pipeline of policing are officers’ daily interactions with the community that are now routinely captured on BWCs. The transcripts of BWC footage combined with NLP and LLMs have the ability to transform this vast trove of data into knowledge. The research into the intersection of AI and BWCs is quickly evolving. We know that evaluating BWC footage using AI is similar to human reviews of the same footage. We also know that officers who use this tool can use that feedback to improve performance. They are also receptive to the AI-derived assistance.

 

 

AI can be used to automate police report writing using BWC transcripts. This automation has the potential to save officers time and increase the readability of their reports. It also has the potential to allow them to spend more time in the community. The only peer-reviewed research on this technology thus far showed no time savings when creating reports. However, the application is new and still improving, and further research is needed. Simultaneously, the development of new AI tools continues to expand. This makes it difficult for researchers and officers to understand all the implications of the new technology.

So, where does this leave police leaders looking for AI guidance, given its unrelenting reach into our lives?

Ideally, police use of technology should be informed by research and best practices. However, in the absence of it, police leaders should be guided by practical knowledge and professional ethics. AI will undoubtedly raise new ethical concerns about its use and application. For example, can officers use an AI-generated summary of a criminal record to inform an arrest decision? If so, in what circumstances?

Using internal agency experts with deep practical knowledge in the field can help answer those questions. A comprehensive internal understanding of agency policies, operational practices, technological infrastructure, and procedures for integrating new technologies is essential and cannot be overstated. Empowering staff to test AI integration efforts through data-driven strategies will generate new knowledge. Their experience in AI’s capabilities will allow them to weigh in on the critical questions regarding its use and application.

In a rapidly evolving world, internal expertise combined with professional ethics can provide the needed knowledge to implement this technology effectively.

 

Watch now

 

The bottom line: Innovation may mean implementing technology, ideas, or practices based on the best available information and professional ethics.

 

Need assistance?

If your agency is exploring the implementation of technology solutions, including artificial intelligence, NPI offers specialized expertise to support a successful deployment and ensure outcomes align with your objectives. For additional information, please complete the contact form below.