Millions of people now take it for granted that Netflix knows more about their specific tastes in gory revenge thrillers than they do themselves. When a woman starts buying unscented lotion and cotton balls, Target's analytics team will ensure she receives ads targeted at pregnant women.
Perhaps inspired by such feats of Holmesian deduction, police departments across the country are increasingly incorporating predictive algorithms into their law enforcement efforts. According to PredPol, a Santa Cruz, California, software developer that uses data about past crimes to predict, very specifically, the places and times in a given city where future crimes are most likely to occur, more than 60 police departments in the U.S. are now using its product.
In general, cops use new technologies to expand the scope of their power and the force with which they apply it. And with 3D face recognition cameras, drones, sensors, automated license plate readers, and similar innovations becoming more common, there's ample reason to believe we're on the verge of an oppressive new Light Age, where state-sanctioned info-warriors equipped with .44 caliber omniscience police our every move, and possibly even our thoughts.
But while this may be the fate we have in store, a more optimistic counternarrative about the impending panopticon exists as well. Imagine a future where every crime that is committed is time-stamped, geo-tagged, livestreamed, and retweeted in 3D high def to a nation of gig-economy stool pigeons and bounty hunters competing for instant bitcoin rewards. As law enforcement gets democratized, the thin blue line will inevitably grow even thinner. Individual privacy diminishes, but so too do arbitrary stop-and-frisks, heavy-handed interrogations, and unjustified use of lethal force.
Granted, a cop-free future isn't what PredPol is selling. "Our software is a tool that does not replace, but requires, the insights of veteran officers and crime analysts," its website advises. And while PredPol's software has potential appeal to a wide range of city agencies, private businesses, and even safety-conscious individuals, the company currently sells exclusively to police departments.
"What we're predicting here is human event patterning in space and time," says Jeff Brantingham, a University of California, Los Angeles, anthropology professor who is one of the researchers whose work ultimately evolved into the PredPol software. According to Brantingham, PredPol's algorithm uses only three data points to make its predictions—the type of crime, the crime's location, and the time of day when it occurred.
Equipped with as much as 10 years' worth of such data, PredPol evaluates it in light of history and certain basic rules. For example, if a crime occurs in a given location, similar crimes are likely to follow shortly thereafter, either in the exact same location or very nearby. Certain locations, like high schools or bars, may experience higher than average crime rates, and some locations may experience more crime during certain seasons or times of day. Essentially, predictive policing expands the hotspot policing that police departments have been practicing for more than two decades now. But instead of merely identifying, say, the handful of streets where 60 percent of all robberies in a given city take place, the software aims to predict what crimes will happen and where on any day.
With PredPol, each police officer is given a map of his city divided into 500-foot by 500-foot boxes at the start of every shift. PredPol highlights the boxes where, based on past patterns, it predicts crimes are most likely to happen that day. The officers, in turn, are encouraged to patrol these areas when they're not responding to service calls or performing other duties.
While many police departments using PredPol have credited it with drops in crime rates, few have engaged in the sort of randomized control trials that would better prove or disprove the software's efficacy. (Brantingham says a study he and his colleagues authored that draws upon a randomized control trial conducted by the Los Angeles Police Department is set to appear in an upcoming issue of the Journal of the American Statistical Association.)
If crime drops statistically attributable to PredPol are occurring, though, it's not because police officers are suddenly breaking up scores of crimes in progress. "Most people think of the police as being in the right place at the right time to catch the bad guy in the act," Brantingham says. "But that is really rare. To the extent that policing impacts crime, it has to be through a deterrence mechanism."
Thus, the general premise that animates PredPol and other predictive policing software: Determine where crimes are likeliest to happen; focus resources in those places; and discourage the acts from even taking place.
But what if sworn police officers aren't the only or even the most effective resource to deploy in such situations? Maybe drones, robots, or citizens paying heightened but remote attention through fixed camera feeds might do the trick. Local business owners with access to PredPol's data could hire private security guards on days when their 500-by-500 section of the city appears on PredPol's hot list. Citizens aware of such data could take special care to avoid these areas, or show up in them en masse as members of ad hoc vigilante flash mobs.
Stretched thin by budget cuts and hiring freezes, police departments are already conducting experiments like this. According to Forbes, the cops in Modesto, California, reportedly park an armored truck equipped with four live camera feeds in one of the PredPol boxes each day. The Los Angeles Police Department's Pacific Division has publicized PredPol hotspots on its Facebook page. "You can simply walk with a neighbor, exercise, or walk your dog in these areas and your presence alone can assist in deterring would be criminals from committing crime in your neighborhood," it advised in March 2014.
"We're not really in the business of telling police departments how to use predictions," Brantingham says. "They already have all the tactics. Some use PredPol to direct non-police resources—drug intervention workers or even street clergy."
In other words, predictive policing might just as easily be called predictive social work or predictive evangelism. In the end, PredPol is simply spitting out data, which could be used to different effect by different people.
What seems most certain at this point is that access to such data is expanding. But it's not just the all-knowing state that has access to it. In Atlanta, a trio of teen sisters created an app called Five-O that aims to be a Yelp or RateMyTeacher.com for cops. In India, a map-based smartphone app called Safetipin encourages users to "record places, harassments, hazards, and audits in a fun way!" In February 2015, Uber announced that it was going to start helping Safetipin in its data collection efforts by mounting smartphones to car exteriors and taking photos at night.
Robocopp, an Oakland, California, startup, is developing a wearable wrist device that will sound a 120-decibel "security siren" and send your precise GPS coordinates to a dispatcher at the touch of a button; the dispatcher will then contact police and also "mobilize an independently contracted professional search team to come looking for you."
Such services will no doubt eventually incorporate predictive policing algorithms of their own, to better alert users to potential problems and more efficiently deploy those Uber-like search teams.
Brantingham is adamant that PredPol only uses data related to space and time, not individual perpetrators. "We don't work off any information about suspects, arrests, anything like that." But how long will it be before others combine predictive policing with face recognition capabilities, mugshot databases, and other criminal records resources to create consumer apps that can alert users to who or what in their vicinity represents the greatest potential threat?
That sort of omniscience, combined with ubiquitous cameras capturing events in real time and thus making long-term evasion extremely unlikely, will lead to major reductions in both privacy and crime. It will also make possible substantial reductions in law enforcement personnel. Not only will Orwell's Thought Police be less likely to emerge in tomorrow's unusually safe dystopia—in the face of peer-to-peer law enforcement, plain old cops are going to grow scarcer as well. They may still loom large in premium cable dramas and video games, but on the well-behaved streets of hyper-surveilled cities, they may soon find themselves scrambling for relevance, just like travel agents, journalists, music executives, and letter carriers.