Predictive Policing and Software Freedom: Should the Police Use Open Source?
Open source software is now widespread almost everywhere, from servers to smartphones. But there's one place where open source is still not common: The world of policing.
Open source software is now widespread almost everywhere, from servers to smartphones. But there's one place where open source is still not common: The world of policing. Here's why open source and software freedom matter so much — and deserve more attention — in the realm of predictive policing.
Like virtually every other type of profession and industry, policing has been transformed by software over the past decade. Police now rely on software tools that use machine learning and data analytics to predict where and when crime will occur. The strategy is called predictive policing.
In one sense, this is not a radically new idea. Fighting crime has always involved analyzing data in order to understand and anticipate criminal activity.
Yet there is a big difference between traditional methods of analysis and modern policing software. Using predictive policing software, crime prediction can occur on a massive — and massively impersonal — scale. Machine learning and data analytics remove the human element from policing, which existed when police predicted and interpreted crime manually.
Predictive Policing Software and the Stakes of Open Source
The free software movement was conceived to ensure that computers users would have control over the code that they ran. By providing access to source code, free and open source software programs allow anyone to study and modify the programs.
The use of software for predictive policing creates an imperative for keeping source code open that extends beyond the traditional goals of free software. When policing and software converge, the stakes are about more than just end users' ability to study and modify software. They're about personal freedom and civil liberty for anyone impacted by the software — which, in the case of predictive policing, means everyone everywhere.
You might think that when the police and judicial authorities rely on software to identify criminals and send people to jail, the source code of the software would be open. That way, anyone could inspect how the program makes decisions about criminality in order to make sure the decision are made fairly and without bias.
But the code of most predictive policing programs is not open source. And as the New York Times reported recently, people are going to jail based on recommendations made by the proprietary code inside these applications.
Should Predictive Policing Software be Open Source?
There are arguments in favor of using proprietary software for predictive policing.
One argument is that these programs are useful and take a lot of time and money to develop. Without the incentive of being able to sell proprietary software, the companies that make predictive policing programs might not produce them.
This sounds like a reasonable argument, but it's actually weak. Lots of companies contribute to popular open source software projects — and many of them generate revenue from it — even though the code is open. In other words, making code proprietary is not the only way to produce good software for predictive policing or anything else.
You could also argue that using software to predict criminality is fairer than relying on humans. Humans tend to be naturally biased, whereas computers don't exhibit prejudice unless they are programmed to do so.
That may be true. But this reasoning doesn't explain why predictive policing software needs to be closed-source. In fact, if you are worried about human bias when it comes to predicting who is going to commit a crime, then you should want to require that predictive policing programs are open source, so that you can make sure human programmers did not insert bias into the algorithms of the programs.
Despite all of the above, most predictive policing software is closed-source. This will probably remain the case until there is greater discussion of the importance of using open source code for predictive policing. So far, this is a niche to which free and open source software advocates have paid surprisingly little attention.
There are signs that things are changing. Earlier this spring, one predictive policing company, CivicScape, open sourced its code. That was a step in a positive direction.
But again, moves like the one by CivicScape remain the exception. In the context of predictive policing, software freedom takes on a whole new level of urgency. It's time for the free and open source software community to pay greater attention.