Guest post by Shawn Parks, Market Manager, Predictive and Business Intelligence Solution Marketing, IBM.
Follow Shawn on twitter: @smparks_mktg
I want to dispel one of the biggest myths I run into when I tell people I’ve worked on projects with law enforcement agencies for predictive policing. That is the image of an officer, in full SWAT gear, breaking down your door to haul you away, throw you in a cell to await trail and be convicted of a crime that you did not commit.
An example takes place in the novel Little Brother, by Cory Doctorow, which explores the paradox of false positives, using an algorithm that is 99% accurate, to catch terrorists in New York. So, if you wanted to catch ten terrorists in a city of 20 million inhabitants, you would need to capture and investigate 200,000 innocent people. Of course, the story is fundamentally flawed by assuming the analytic output is being considered in a bubble. A police department (or any organization) that acts solely on analytics and not in the context of situational awareness is set up for failure.
First, using predictive models to deploy officers to a given location or to assist crime analysts in generating leads for cold cases is very different from claiming to be able to prevent another 9/11 or Boston Marathon bombing. It’s like fixing a dripping faucet instead of looking for the pipe that’s about to burst. Second, Doctorow points out that the software is not really 99% accurate, which is true. But that leaves us to question the accuracy rating of its competitor, the human mind. We regularly make mistakes or errors in judgment, or simply forget things. Yes, algorithms, math and software can deliver inaccurate results when working with inaccurate data. But I will take the 60% accuracy of a computer program crunching out patterns and correlations in data points over the human brain any day. Just look at Watson.
Lieutenant Arnold Palmer of the Miami-Dade Police Department made an interesting point awhile ago about not just making more arrests, but more convictions. In other words, he believes it’s important that police departments work both more efficiently and more accurately. It’s an interesting concept that’s more difficult to put into practice than it sounds.
If you’re looking over your shoulder, worried about a wrongful arrest charge, fear not. A case that is based on “my computer told me to do it” will hold up in court about as well as “my dog ate my homework” did in third grade. Intelligence-led policing does not replace good police work. Instead, it supplements action that typically relies on experience and gut feel.
Police departments can benefit significantly from using historical data to transform the way they manage their “business.” That’s especially true for those devoting resources to deploying officers—often putting them in harm’s way—without all the tools at their disposal.
See how other police departments are using IBM predictive policing solutions
IBMers, especially those of us on the Business Analytics team, work with the concept of using data to drive decision-making on a daily basis. We may forget that operationalizing this concept takes careful planning, strong leadership and focus on the overall value of the project. However you look at it, we are not changing what police departments do at the core. We are changing how they do it, and using technology to help them do their jobs better.
Watch this video showing how the Miami-Dade Police Department uses analytics.
Learn more about predictive policing here.