How Artificial Intelligence is Changing Law Enforcement

Written by Scott Wilson

robotic dog

If you want court-side seats to one of the biggest debates over artificial intelligence, then watching how law enforcement is approaching the technology is the place to be.

Facial recognition, driven by computer vision systems, is already making a splash, with the FBI assembling a database of more than 640 million photos to match against. With body worn cameras becoming standard in more and more departments, those numbers will only rise. Fingerprint matching is another place where computer vision is making a mark, with less controversy.

Law enforcement is already dipping its toes into the AI waters.

The first waves of predictive analytics driven by AI have also drawn fire from civil liberties and community groups. By using ML algorithms to map crime hotspots, police can alter their own resource allocation and patrol patterns to match. But such high-stakes bets risk reinforcing biases if the source data and algorithms aren’t carefully vetted.

It’s a challenging area for artificial intelligence professionals to work in. But if the alignment problems in law enforcement can be addressed fairly and effectively, that work will be a bright spot for AI development in general. This is a field with a lot riding on it for both the future of law enforcement and the future of AI.

How Artificial Intelligence Is Poised to Play a Big Role in Law Enforcement

police robotsWhen you are talking about lives and livelihood, extra care and attention to detail are key.

A criminal accusation, let alone a conviction in a court of law, has big consequences. Police know better than anyone that they have to get it right every time.

AI is attractive to many in the law enforcement and public policy communities because it has the potential to take the variability out of legal matters. An algorithm doesn’t have a bad day or a short temper. The hungry judge effect, where studies have found that jurists are more likely to deliver harsh sentences in similar cases just before meal breaks, can apply to cops too. But society has a right to expect officers to have an even hand, even on a bad day.

Tackling the Hard Problem of Bias in Law Enforcement AI

ai ethics folderEarly deployments of AI have sometimes reflected the biases of society toward minority groups. And some of the explorations of those accusations have revealed an apparently intractable problem: it’s mathematically impossible to satisfy different definitions of fairness.

An algorithm called COMPAS, the Correctional Offender Management Profiling for Alternative Sanctions) has been used for years to provide sentencing recommendations, was the subject of an investigation by journalists in 2016. They found that while it correctly estimated recidivism in prisoners, it was more likely to place Black criminals in a higher risk category than where they belonged.

This is the hard problem for artificial intelligence professionals working on law enforcement applications. More even than the technical details, they’ll have to understand various statistical realities and social justice demands just as clearly as they do algorithms and training methodologies.

Careful AI Deployments Are Coming to Every Facet of Law Enforcement

One thing is certain about AI in law enforcement: it’s coming.

In addition to tools like COMPAS, the sentencing risk assessment software that created such a furor over racial discrimination and revealed the impossibility of matching predictivity and equality, police are already rolling out AI tools like:

There are also many administrative benefits that AI can bring to policing. Officers spend, on average, around three hours per shift on report writing and paperwork. Basic natural language processing tools have the potential to reduce that burden considerably. AI can draft, review, and streamline much of the paperwork that powers American justice.

You may be finding that a chatbot is your next 911 operator.

Computer-aided dispatch already exists, but AI tools can do a better and faster job of categorizing calls, allocating resources, and finding relevant information to assist officers responding.

Like every sizable organization, the police have plenty of basic administrative tasks to fulfill. Record-keeping in law enforcement takes on a little more weight than it does for your typical manufacturing business, though. Careful chain-of-evidence records and careful connections between criminal violations are essential to justice. AI can track entries, make those connections, and ensure data entry is complete at every step.

Other AI tools are aimed squarely at criticisms of bias: San Francisco is planning an AI-enabled system to automatically redact race and other characteristics that may bias prosecutors from the police statements they use to make charging decisions.

Law enforcement will need to tap expertise who can help them deal with the onslaught of criminal uses of artificial intelligence, as well.

AI will also play a role on the tactical side of law enforcement. Advances in robotics and drone systems are giving police the opportunity to engage with dangerous situations or subjects without putting officers in harm’s way. The more capable and independent those devices are, the wider the span of situations where they can be used.

Further, the advanced logic and reasoning abilities of AI could be used in detective work. Finding patterns that aren’t obvious to humans is a key feature of machine learning systems already. In the future, connections between crimes could be surfaced by smart systems fed mountains of evidence.

Naturally, this requires extraordinary consideration of social and ethical concerns. While many of the big decisions will be made well above the pay grade of the humble AI engineers putting these systems together, it’s up to AI professionals to understand the implications. Input on the true reasoning capabilities and the potential biases of AI law enforcement systems will inform those big decisions.

Where Will the Jobs Be for AI Professionals in Law Enforcement?

examining hologramMost police forces, except for the very largest, are not likely to be hiring AI engineers or trainers themselves to build and manage these tools. Instead, most jobs in this field will be with various vendors who build tools and software for government and police agencies.

An exception may be at the federal and state level, where AI experts may be needed to evaluate and ensure quality and justice standards in the use of these tools.

In any case, the full range of various AI positions will be needed:

Artificial Intelligence Degree Programs Prepare Students for Roles in Law Enforcement Development

law enforcement classroomThe kind of degrees needed to land those jobs are fairly standard. There are no specialized courses of study for AI applications in law enforcement.

Instead, a typical Master of Science in Artificial Intelligence or a Bachelor of Science in Artificial Intelligence and Machine Learning provide all the basics for building law enforcement tools with AI and ML. From high level mathematics and statistical skills to the hands-on world of coding and algorithm development, these programs cover all the tools.

More critical, high-level work in law enforcement AI will fall to those with graduate degrees, while more day-to-day maintenance and training will be open to those with bachelor’s degrees.

For experts who are already working in IT on law enforcement projects, a Certificate in Artificial Intelligence may be the right bridge to jump to AI positions. While these skip a lot of the more comprehensive studies in a full degree, they are much less expensive and faster to earn. For students with sufficient background in the core technical skills, they can deliver similar qualifications to a full degree.

Professional certification, a process by which independent organizations validate skills in specific areas, is not a big feature in law enforcement AI work. There are no certs that are specific to the field. However, various professional certifications in artificial intelligence may help illustrate overall qualifications as a machine learning or AI engineer.

Any kind of education aimed at preparing professionals for law enforcement AI development roles will have to put ethics front and center. Although the ultimate decisions made in each case will still always come down to judges and juries, there’s a lot that happens in the system before it gets that far. AI professionals working within the field will need to ensure that fairness and quality are baked in at every stage of that process.