Using artificial intelligence to improve efficiency before, during and after a call
AI in law enforcement offers potentially huge benefits in improving officer efficiency and community safety
This feature is part of our Police1 Digital Edition, a supplement to Police1.com that brings a sharpened focus to some of the most challenging topics facing police chiefs and police officers everywhere. To read all of the articles included in this issue, click here.
By Ron LaPedis
Everyone seems to have an opinion about what artificial intelligence (AI) means, with many peoples’ opinions fueled by movies and TV representations of AI taking over the world. The reality is much simpler than that. AI is software that can analyze huge amounts of data to detect patterns that it would take humans much longer to see. This software can evaluate free-form text, time and location information, audio, photos and videos.
ARTIFICIAL INTELLIGENCE IN LAW ENFORCEMENT: A DAY IN THE LIFE
To understand the breadth of applications in policing, let’s review the key areas AI can have an impact in improving officer efficiency and community safety.
1. Before officers are dispatched
Several calls come in to dispatch about a woman screaming like she is being attacked.
AI software can pull location data from the CAD system to determine if the calls surround a specific area, then drill down into locations in that area that are already known, such as a drug house, bordello or where domestic violence has been reported before. Let’s assume that there are hits on a few locations, but one stands out with a prior domestic violence report.
2. During response to a call
While officers are headed to the call, the dispatcher uses the AI software to pull the details of past incident reports, firearms ownership records and arrest records for the occupants. If weapons charges show up, dispatch can let the responding officers know to take appropriate precautions while additional support is being dispatched, possibly to include the SWAT team. Officers also know if the suspect might be considered more dangerous than average.
If a suspect is taken into custody, the officers already have his rap sheet and can proceed appropriately. Perhaps the perp is a known gang member and the incident allows for an extensive search of the premises.
3. After the call
If a gun comes in as evidence from the call, an officer can ask the AI system to bring up all reports that may be relevant to that specific firearm while filtering out reports that would just clutter up the case. Similarly, if drugs are found packaged in a specific manner, that may also be a “signature.”
The use of AI-based data mining is much faster than going through filing cabinets pulling reports or even manually searching through computerized records. Since the perp is a known gang member, you may have lucked out and have found a firearm that can be linked to other crimes. This may seem like magic, but we’ll talk about how you can help your AI help you in the next section.
GARBAGE IN, GARBAGE OUT
Computer geeks know that the help computers can provide is only as good as the software and data that they have to work with.
Let’s say that officers are picking up a lot of 9mm shell casings all over town. Is there anything special about those shell casings? Are they from a specific manufacturer? Are there firing pin, extractor or ejector marks on them? What color is the sealant around the primer? By entering all of this information into your records management system (RMS), you give your AI software a much better chance at selecting the right records and eliminating information irrelevant to your current investigation. If you enter as many details as possible, your AI may be able to tell you that three different firearms are being used, and that each one is showing up in a specific set of locations. Ammo with unique characteristics might be able to be linked back to a specific vendor. Maybe one or more gangs are sourcing reloaded ammo from the same person.
Without AI, this work could day days or weeks; but with AI, you could have your report in minutes. When a firearm shows up and is test fired, AI can give you the ability to link it to existing records, and if you have made an arrest along with it, you can link its owner to one or more crimes while they are in the interrogation room.
USE WHAT WORKS
Natural language, pre-formatted text and pre-identified images are the meat and potatoes of AI. By using these tested and well-understood components of an AI solution, LE can rely on the simplest and most mature AI solutions to ease the learning burden on officers and ensure that the results will be more predictable and reliable than trying to force leading-edge solutions into work that it may not be ready to do.
Even without facial or image recognition, images and video can become a part of your big data by having officers or clerks appropriately tag them from their daily reports. Specific sections of video can be tagged by including the timecode in multiple description records which the AI can search. Descriptions can include contact names, geographic location and any other items of interest in the clip.
Because AI is evolving so rapidly, you will want to read up on it and track down tech-savvy staff for help. If you don’t believe that you have the expertise needed to properly evaluate an AI system, then it is time to seek out consultants. AI is used across so many fields and a consultant in one field may not know the subtleties that law enforcement requires. While you could start with a non-LE consultant to help you understand what AI can and cannot so, when it’s time to start writing policies and specifications, you should seek out a LE-specific consultant.
Here are some terms commonly used when discussing AI:
- Big data: Extremely large data sets that may be analyzed (mined) to reveal patterns, trends and associations, especially relating to human behavior and interactions.
- Machine intelligence, machine learning or deep learning: Intelligence demonstrated by machines in contrast to the natural intelligence displayed by humans and animals. This includes reasoning, knowledge representation, planning, learning and natural language processing.
- Natural language (free-form) processing: Programming that lets computers understand, interpret and manipulate human language rather than data that has been pre-formatted.
About the author
Ron LaPedis has been a business continuity and security professional for over 25 years and frequently writes and speaks on business continuity, cybersecurity, physical security and public safety topics.
Ron is a Master Business Continuity Professional (MBCP), a Fellow of the Business Continuity Institute (AFBCI), a Distinguished Fellow of the Ponemon Institute and a Certified Information Systems Security Professional (CISSP).