Posted by Elizabeth Joh on Wed Nov 06 2013
Policing has jumped on the big data trend. What is big data? Apart from being a catchy moniker, big data refers to an important development in fields as diverse as internet commerce, public health, transportation management, language translation, and even presidential politics. Nate Silver’s number crunching of the 2012 election, Google’s flu trends, and Amazon.com’s suggestions regarding “customers who bought this item also bought” are examples of big data. While there isn’t a universally agreed upon definition, the term “big data” typically refers to the application of artificial intelligence—specifically predictive analytic software—to vast quantities of data.
Vast is somewhat of an understatement. By one estimate, in 2012 nearly all of the world’s information constituted about 2.7 zettabyes: that’s 2.7 followed by 21 zeros. Computer analysis of those numbers helps us see the world in new ways that simply weren’t possible when analysis relied on people alone.
One of the most talked about uses of big data in law enforcement has been the adoption of predictive policing. This involves the application of predictive analytics to crime data in order to forecast criminal activity. Sound straight out of Minority Report? Perhaps (minus the bathtub-dwelling triplets). But while Tom Cruise’s character identified which persons would likely commit future crimes, predictive policing software tries to isolate small geographic areas where crime is likely to occur. Research has shown repeatedly that crime doesn’t occur at random. Instead, crime tends be concentrated in limited areas. Predictive policing software tries to identify these areas with precision.
In 2011, the Santa Cruz Police Department took predictive policing to the streets. The department experimented with a program that relied upon five years of crime data and applied to it software modeled on earthquake aftershock theory. Before heading out on their shifts, police officers could consult the resulting maps that identified 500 by 500 feet blocks within the city likely to be targeted by crime. (The police were encouraged, but not required, to pay attention to these areas.) The Santa Cruz Police Department attributes a significant drop in burglaries to its experimental adoption of the program. Other cities such as Los Angeles and Seattle are following suit.
As someone who thinks about the regulation of policing, I wonder whether such tactics are constitutional. The short answer is that we don’t know yet, although criminal defendants almost certainly will challenge the use of these techniques sooner or later as Fourth Amendment violations. Investigative stops require at least reasonable suspicion, and the argument may be that predictive software cannot be the basis of that justification.
How courts will resolve this question will likely turn on two issues. The first is the degree to which the police actually rely on these predictive models. For now, it appears that the police use the predictive maps to determine which areas of a city might need extra attention. To reach the point of stopping a person for questioning or further investigation, the predictive models aren’t by themselves telling the police which persons deserve further scrutiny. Perhaps computer modeling will become so sophisticated that the level of human judgment might be small. That day, if it comes, may upend what we think about terms like “reasonable suspicion.”
The second issue involves whether predictive policing is like other situations in which the police have used information to justify investigative stops of persons. The Supreme Court’s Fourth Amendment decisions have permitted the police to rely upon tips, even anonymous ones, to provide part of the justification needed for such stops. Is a computer algorithm like an anonymous tipster? In some sense, a piece of software may be superior to a tipster, who may harbor a grudge or other questionable motives.
While predictive policing is rolling out slowly now, watch for it to spread, and expect thorny legal questions to emerge.
Elizabeth E. Joh is a Professor of Law at the University of California, Davis, School of Law. She is working on a longer article on the uses of big data by the police.