Crime Prediction Keeps Society Stuck in the Past

One of the most remarkable examples of the use of predictive technology is the story of Robert McDaniel, detailed by journalist Matt Stroud in the Verge in May 2021. McDaniel is a resident of Austin, a Chicago neighborhood that saw 72 homicides, nearly 10 percent of the city’s total, in 2020 alone. Despite the fact that McDaniel had no record of violence (he had been arrested for selling pot and shooting dice), a Chicago Police Department predictive policing program determined in 2013 that he was a “person of interest”—literally. In the 2011-16 CBS crime drama of that name, “the machine,” created by the show’s protagonist, can only determine that a person will be either the victim or the perpetrator of a violent crime, but not which. Similarly, the algorithm used by the CPD indicated thatMcDaniel was more likely than 99.9 percent of Chicago’s population to be involved in a shooting, though which side of the weapon he’d be on was unknown.

Equipped with this “knowledge,” Chicago police officers placed McDaniel on their Strategic Subject List, later known as the “heat list,” and kept a close watch on him, despite his not being under suspicion of involvement in any specific crime. Because some of that surveillance was overt, it suggested to others in his neighborhood that he might have some kind of connection to the police—that he was perhaps an informant, a tremendously damaging reputation. 

Predictably enough, McDaniel has been shot twice since he was first identified by the CPD: first in 2017, perhaps partly due to publicity generated by his appearance that year in a German documentary, Pre-Crime, that he hoped would help to clear his name; and more recently in 2020. He told the Verge that both shootings were due to the CPD surveillance itself, and the resulting suspicion that he was cooperating with law enforcement. “In McDaniel’s view,” Stroud writes, “the heat list caused the harm its creators hoped to avoid: It predicted a shooting that wouldn’t have happened if it hadn’t predicted the shooting.”

That is true enough, but there is a deeper pattern to observe here as well. Because of police data from the past, McDaniel’s neighborhood, and therefore the people in it, were labeled as violent. The program then said that the future would be the same—that is, that there would not be a future, but merely reiterations of the past, more or less identical with it. This is not merely a self-fulfilling prophecy, though it certainly is that: It is a system designed to bring the past into the future, and thereby prevent the world from changing.

The program that identified McDaniel appears to have been developed specifically for CPD by an engineer at the Illinois Institute of Technology, according to earlier reporting by Stroud. The CPD program identified around 400 individuals most likely to be involved in violent crime and put them on its heat list. That program started in 2012 and was discontinued in 2019, as disclosed that year in a Chicago city government watchdog report that raised concerns about it, including the accuracy of its findings and its policies concerning the sharing of data with other agencies. The custom CPD algorithm reportedly focused on individuals, and it likely resembles a wide range of programs used by law enforcement and militaries of which the public has little knowledge. For instance, in 2018, journalist Ali Winston reported in the Verge that the surveillance company Palantir, founded by Peter Thiel, had been secretly testing similar technology in New Orleans since 2012 without informing many city officials.

Better known by the public are programs like CompStat and PredPol, which differ from the CPD heat list in targeting geographic areas rather than individuals. CompStat was developed by the New York City Police Department as a data-driven approach to policing, where officers gathered crime statistics by district and then used that data to inform police allocation. There is a wide variation in the lore regarding CompStat: It is either responsible for the drop in New York’s crime rate, or it had no meaningful effect on the amount of crime and simply contributed to more racist policing, depending on whom you ask. 

PredPol, meanwhile, is more predictive. (The software behind the widely used platform has roots in predicting battlefield casualties in Iraq.) Even so, it operates from the central premise that by using historical crime data—notably crime type, location, and time of offense—the proprietary algorithm can predict where future crimes are likely to occur. In an analysis of a trove of PredPol data left available on the open web, Gizmodo found that the system “relentlessly targeted” areas predominantly made up of people of color and the poor.

All of these policing systems operate on the assumption that the past determines the future. In Discriminating Data: Correlation, Neighborhoods, and the New Politics of Recognition, digital media scholar Wendy Hui Kyong Chun argues that the most common methods used by technologies such as PredPol and Chicago’s heat list to make predictions do nothing of the sort. Rather than anticipating what might happen out of the myriad and unknowable possibilities on which the very idea of a future depends, machine learning and other AI-based methods of statistical correlation “restrict the future to the past.” In other words, these systems prevent the future in order to “predict” it—they ensure that the future will be just the same as the past was.

“If the captured and curated past is racist and sexist,” Chun writes, “these algorithms and models will only be verified as correct if they make sexist and racist predictions.” This is partly a description of the familiar garbage-in/garbage-out problem with all data analytics, but it’s something more: Ironically, the putatively “unbiased” technology sold to us by promoters is said to “work” precisely when it tells us that what is contingent in history is in fact inevitable and immutable. Rather than helping us to manage social problems like racism as we move forward, as the McDaniel case shows in microcosm, these systems demand that society not change, that things that we should try to fix instead must stay exactly as they are.

It’s a rather glaring observation that predictive policing tools are rarely if ever (with the possible exception of the parody “White Collar Crime Risk Zone” project) focused on wage theft or various white collar crimes, even though the dollar amounts of those types of offenses far outstrip property crimes in terms of dollar value by several orders of magnitude. This gap exists because of how crime exists in the popular imagination. For instance, news reports in recent weeks bludgeoned readers with reports of a so-called “crime wave” of shoplifting at high-end stores. Yet just this past February, Amazon agreed to pay regulators a whopping $61.7 million, the amount the FTC says the company shorted drivers in a two-and-a-half-year period. That story received a fraction of the coverage, and aside from the fine, there will be no additional charges.

The algorithmic crystal ball that promises to predict and forestall future crimes works from a fixed notion of what a criminal is, where crimes occur, and how they are prosecuted (if at all). Those parameters depend entirely on the power structure empowered to formulate them—and very often the explicit goal of those structures is to maintain existing racial and wealth hierarchies. This is the same set of carceral logics that allow the placement of children into gang databases, or the development of a computational tool to forecast which children will become criminals. The process of predicting the lives of children is about cementing existing realities rather than changing them. Entering children into a carceral ranking system is in itself an act of violence, but as in the case of McDaniel, it also nearly guarantees that the system that sees them as potential criminals will continue to enact violence on them throughout their lifetimes.

A widely popular and oft repeated claim about algorithms and “artificial intelligence” is that, given enough data over a long enough period of time, the algorithm can not only deliver to you what you desire, but that it can do so before you even desire it—that, in effect, that algorithm knows you better than you know yourself. We see this claim wherever AI is at work, whether that is a Spotify playlist, your Amazon wish list, or your Netflix movie choices. So, in the case of algorithms that claim to know you’ll commit a crime before you do, it’s worth asking the question: What does a racist and carceral society desire? Without question, many in such a society—one that incarcerates more human beings, many of them Black and brown, than any other place on the planet—desire to maintain the status quo.

In the case of all these algorithms, what they typically deliver is not a new experience, but extra helpings of what you’ve had in the past. They don’t anticipate your desires as much as they assume that past desires and future ones are mostly similar. In the case of a musical playlist, the stakes are small. In the case of anticipating the likelihood that someone is involved in a shooting, or locking people in cages—not so much. But until there is a radical shift in how we think about “crime,” policing, technology, and the ways they intersect, the future of predicting the future of crime is destined to promise more of the same.