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.
The idea of intelligent machines has long been the subject of science fiction films. Fortunately, we now live in an era in which Artificial Intelligence (AI) and Machine Learning (ML) is no longer fiction, but reality.
The terms AI and ML are heavily used in the private sector, especially among companies involved in Big Data, analytics, information technology and process improvement. Even though the terms are often used interchangeably, the differences are important to understand.
Artificial Intelligence (AI) is a much broader term related to how a machine “thinks” or processes information. Think of AI as the brain of a person. Machine learning (ML) however, is related to the process of how a machine learns. Machines learn in different ways – just as people do.
AI and ML technologies are extremely powerful and efficient ways to process large amounts of information. There is no question these technologies are becoming a permanent part of the law enforcement ecosystem.
Here are a few considerations to keep in mind before purchasing AI and ML technology.
USE HUMANS AND AI; NOT ONE OR THE OTHER
A crucial component of the proper application of AI and ML technology is keeping the human component. Many ethical concerns related to using these technologies stem from removing the human component of AI – and rightfully so.
Bud Levin, an expert in AI at the FBI’s Training Division, reiterates the fact that taking the human element out of AI is potentially dangerous since machines do not understand the “complexities and absurdities of the Constitution, our laws, and policies.” [1]
The goal of any good AI and ML program is to encourage officers to do what they normally do but to increase their efficiency and effectiveness to better protect society. AI and ML should always be used in conjunction with humans, not as a replacement for, humans.
THE NEED FOR GOOD DATA
AI and ML technologies are progressing rapidly, and for good reason – they are learning. It is now possible to purchase “mature” machines that have already made mistakes in their learning rather than relying on adolescent machines. Sounds pretty close to parenting, right. Just as kids are more prone to make mistakes, adolescent machines are also more prone to make mistakes.
This is where the difference between AI and ML comes in. Machine learning, especially deep machine learning, means that instead of being programmed to perform specific tasks, a computer can learn independently and adapt to its understanding based on the data it receives. The more “good” data it is exposed to, the better the computer can begin to recognize and identify objects based on complex pattern recognition. “Bad” data in ML can lead to mistakes.
Agencies must start now to develop good data entry habits for their officers.
The reason for good data is simple:
Good data in = Good information out
Bad data in = bad information out
The entire foundation of ML is data. It learns by using the data it is given. Machines, just as humans, can make mistakes by relying on misinformation, false information, or inaccurate information. For that reason, good data collection is a critical role of good ML application.
Responsible Application of AI and ML
There are many ethics-related questions surrounding AI and ML use in law enforcement. Agencies need to consider how and why they are going to use these powerful technologies. More importantly, agencies need to document in detail the how and why in a formal departmental policy.
Luckily, agencies do not have to start fresh when developing policy and implementation guidelines for AI and ML technologies. The London based, The Alan Turning Institute, was created in 2015 to provide data science guidance to both the public and private sectors. In 2017, as a result of a government recommendation, they added guidance and the proper application of artificial intelligence in a joint publication with London’s Informational Commissioners Office (ICO).
This publication can be used to easily develop departmental policies.
REAL-LIFE APPLICATIONS OF AI AND ML
Agencies are drowning in data, collecting terabytes of it each day. AI and ML’s primary use is data management, specifically making large amounts of data searchable, filterable and retrievable in real-time.
Here are just some of the current law enforcement applications for AI and ML:
1. Faster video analysis and redaction
Video analysis is a resource-intensive task for law enforcement. Many agencies report spending up to two hours redacting a five-minute video for public release or spending 1:1 time watching video surveillance footage in hopes of finding suspects, missing people, or evidence. Those hours can quickly add up.
Because of AI and ML technology, officers can filter their searches to look for specific characteristics; for example, a man in a red shirt. The system then will look for just the man in a red shirt. Once the man is identified, the officer can redact or remove all non-involved or non-pertinent information leaving just the man in the red shirt on the screen.
2. Audio analysis and transcription
An officer’s radio communications is an important piece of evidence in many criminal trials. But sorting through the hours of audio for nuggets of specific information can quickly become overwhelming, which is why many agencies are stopping that practice. Luckily, AI and ML technology can help filter out what is essential by scanning the data for specific information, like when an officer yells, “He has a gun,” or when a suspect is heard in the background, “I am going to hurt you.”
A bonus to audio analysis is real-time audio transcription. Instead of listening to the audio, the system can print out the full audio transcription of the recording to be used in an officer’s report, or as evidence.
3. Real-time search
Because of audio recognition advancements, AI systems can quickly find information needed by the officer, in real-time. An officer can ask the AI system to run a person’s name and date of birth, for example. The system then can respond with driver’s license information, warrant information, recent arrests, recent involvement, or recent alerts directly to the officer.
4. Gunshot detection
One of the fastest evolving use for AI and ML is in gunshot detection devices. Data science companies and scientists are working to develop advanced algorithms to detect the exact location gunshots, how many rounds were fired, and the number of firearms present. Because of AI technology, these systems can even differentiate between muzzle blasts from shock waves to even determine the type of gun (rifle, shotgun, handgun).
5. Crime forecasting
Many agencies are using AI to help combat crime. These systems can scan through volumes of information to provide accurate crime forecasting using predictive analytics. Agencies can then use that data to allocate police resources.
6. Report Management Systems (RMS)
RMS 3.0 technology incorporates the use of AI technology into report writing. These systems have the capability of searching narrative content by using search engine optimization (SEO)-like algorithms. AI will significantly increase the analyst’s ability to find specific information instead of reading countless reports to find that same information.
7. Forensic analysis
Any good agency approach to AI and ML would not be complete without including applications for the forensic team. AI and ML technology are currently used in DNA analysis. As DNA technology advances, so has the sensitivity of DNA analysis. Forensic scientists can now detect and process low-level, degraded, or otherwise unviable DNA evidence that could not have been used just a few years ago.
There are very few crimes that do not have at least some digital component to them. Human trafficking investigations are improving because of AI technology. Internet companies like PayPal are using AI to detect fraud, and the U.S. Department of Transportation is using AI to track real-time accident, weather, lighting and traffic conditions.
Artificial intelligence and machine learning technology are available now to law enforcement worldwide. Is your agency prepared for the future of policing?
About the author
Joshua Lee is an active-duty police sergeant for the City of Mesa (Arizona) Police Department. Before promoting, Joshua served five years as a patrol officer and six years as a detective with the Organized Crime Section investigating civil asset forfeiture, white-collar financial crime and cryptocurrency crimes. Joshua is a cryptocurrency, money laundering and dark web consultant for banks, financial institutions and accountants throughout Arizona. He also serves as one of Arizona’s subject matter experts on cryptocurrency crimes and money laundering.
Joshua holds a BA in Justice Studies, an MS in Legal Studies and an MA in Professional Writing. He has earned some of law enforcement’s top certifications, including the ACFE’s Certified Fraud Examiners (CFE) and the IAFC’s Certified Cyber Crimes Investigator (CCCI). Joshua is also an adjunct professor at a large national university and smaller regional college teaching, law, criminal justice, government and English courses. He instructs police in-service training and teaches at the regional police academy.
