• Book Review of Predict and Surveil: Data, Discretion, and the Future of Policing, by Sarah Brayne.1
Sarah Brayne has done some excellent sociological research by spending several years embedded in the LAPD [Los Angeles Police Department]-one of the most technologically advanced police departments in the country. By doing so, she has given us a chance to glance behind the curtain of big data in policing just as it is in its ascendency. While she issues some warnings and even possible solutions to the inevitable overreach for which such a shift allows, she also admits that there are some real advantages to be gained in terms of crime prevention (although it’s too early to properly assess the overall trade-offs between privacy and crime prevention).
“Will big data lower crime and reign in police abuse? Or will it exacerbate structural inequalities that already typify our criminal justice system?”

The most striking thing about this excellent and timely book is that it leaves one in a state of healthy ambivalence: Will big data lower crime and reign in police abuse? Or will it exacerbate structural inequalities that already typify our criminal justice system? Predict and Surveil1 will be of great interest to anyone thinking deeply about privacy, policing, and the way either of these interact with the law in this technologically revolutionary moment.

The Way We Live Now

We all have a general sense that we’re leaving digital trails wherever we go, and there’s been much consternation about the use of such data in the marketplace. Brayne argues that we are dealing not just with a giant leap forward in terms of the sheer quantity of data that’s being collected on us, but also in terms of the aggregation of data from multiple sources, which allows individuals to be closely monitored. While police are often legislatively restricted in terms of what data they can collect on citizens, they can now work around most limitations since there are no limits on what data they can buy from private sources. Private companies, like Palantir, that help police aggregate their own data have encouraged this ‘function creep’ in which police become consumers of private data collections as well as of government intelligence originally meant for other purposes. Brayne notes that Palantir itself was “originally designed for counter-insurgency efforts in Iraq and Afghanistan,” and I couldn’t help but think of Coyne and Hall’s Tyranny Comes Home, a close documentation of the migration of military efforts to the control of the domestic population (7).

Brayne is quick to point out benefits, such as in her opening example of a body dump case that was solved in just two days with the help of a license plate reader and the gang-tracking system CalGang. Predictive algorithms that send officers to particular areas (called “red boxes”) to patrol may have a real deterrent- or at least, displacement-effect. Further, data can do as much to exonerate as to convict us, since it can demonstrate that we were nowhere near a crime scene, for instance. Big data also has the potential to finally provide a watcher for the watchmen as well, as police activity can be more closely monitored and analyzed too. However, before the reader is tempted to romanticize the efficiency, precision, and potential social advantages of big data policing, Brayne suggests a few ways it can go very wrong.

Power Imbalances

First, while one would hope that data-driven policing would reduce bias and disproportionate enforcement by removing the ‘human element,’ it can also serve to reinforce and perpetuate these things. The best example she gives is the LASER program and the attendant Chronic Offenders Registry, although we find out at the end of the book that this particular program has been shut down due to public outcry. Police prioritize targets with a point system -five points if someone is in a gang, five for prior convictions, and so on. This allows the more likely offenders to float to the top of consideration when searches return a large number of results. So far, so good. But one point is also added any time someone has contact with the police, even if officers are just having friendly conversations, or ticketing someone for jaywalking or failing to use their turn signal. There’s a strong incentive to fill out field interview cards (FI’s) in order to build data by getting people into the system (64-67). She recounts a few shocking, but quite real, scenarios, as when parolees are visited by police three or four times in one day, or “ghetto-birds” (police helicopters) fly over a neighborhood 80-90 times a week. It’s not hard to imagine how a person from a rough neighborhood could start racking up points without any criminal activity at all. Since high-crime neighborhoods are far more likely to be under surveillance, people are more likely to be caught for trivial things, even though only a small percentage of the neighbors are involved in the sorts of dangerous activities that we’re really interested in addressing. The higher one appears on the points list, the more likely one is to be surveilled, leading to more interactions with the police, more points, and so on.

In contrast to this snowballing of police interaction and surveillance for marginalized groups, those with more political clout can avoid surveillance. Gun owners have successfully avoided a federal gun registry, and police themselves actively resist their own surveillance on the job through their police unions. Brayne tells of a ride-along in which she expressed surprise that the officer had to call in his location. When she informed him that she had assumed the cars all had GPS trackers, he responded that they do, but that the police unions threw a fit and so they’ve never been turned on. Brayne does not argue that there should be a federal gun registry or that officers should be tracked everywhere they go. She simply points out that those least able to legally and politically resist being constantly surveilled are also those most likely to be under surveillance.

Data and the Law

A central take-away of the last portion of the book involves the inability of traditional privacy law to deal with the way data works now. Here, Brayne makes her boldest claim: “[l]egal constructs like the third party doctrine are set up in a way that is basically inapplicable to modern life” (131). The third party doctrine holds that government officials can access any information that I voluntarily give out to others without violating my Fourth Amendment rights. But in today’s environment, it’s nigh impossible to communicate with other people at all without “exposing data to a third party,” so that “the third party exception has become the exception that swallows the rule.” Computer-aggregated data is just not the same animal as the data an officer can collect by reading a letter I wrote to a friend.

Furthermore, she highlights the programmatic nature of police surveillance: “ongoing, cumulative, and sometimes suspicionless data collection and use” (129). In response to these new challenges, she considers various academic views, including the suggestion that administrative law, rather than criminal law, may be a more appropriate way to govern this type of surveillance. She cites Orin Kerr’s work on the Fourth Amendment as an example of the hope that regular calibrations through case law will still be sufficient, even in our digital age. She also worries that the exponential increase in plea bargaining and the correspondent loss of trials means that many relevant cases to this issue will never see the inside of a courtroom. What’s worse, so much of the big data story happens before any real evidence is collected, by a kind of case-building process that will remain unseen even if the case itself does come to trial. Police have even learned how to create ‘parallel constructions’, a process of lying about the way that the case actually progressed in order to obscure that a surveillance strategy had anything to do with the course the case took.

Assessing the Trade-Offs

Although police departments talk big, there is no overwhelming evidence that algorithms are serving us better than humans have when it comes to good policing. We certainly see some correlation between big data platform implementations and reductions in crime, but those reductions are often a continuation of a trend that had already begun before such resources were available. Remember that the LAPD is a first adopter, so Brayne is giving us a window into a world that is only just beginning in many cities.

For more on these topics, see the EconTalk episodes Franklin Zimring on When Police Kill and Cathy O’Neil on Weapons of Math Destruction.

Brayne has high hopes for positive uses of data in policing, such as the direction of non-punitive interventions (as in mental health cases) and to clear current cases and solve cold cases. But she’s right to warn us that we do not yet know whether the dangers of big data are outweighed by the possible gains, and we won’t know that till we have more independent research. Her most chilling warning, though, is the legal one. Is the world of big data so different from the old world that the usual adjustments of the common law will be unable to maintain the reality of privacy? And for those of us who want to avoid the expansion of administrative law to address our new reality, what’s the alternative?


[1] Sarah Brayne. Predict and Surveil: Data, Discretion, and the Future of Policing. Oxford University Press, 2020.

* Rachel Ferguson is a Professor of Managerial Philosophy, co-chair of the Lindenwood Honors College, and Director of the Liberty and Ethics Center in the Hammond Institute. Her research interests include Hume’s classical liberalism, the philosophy of economics, and Aristotelian virtue theory.

As an Amazon Associate, Econlib earns from qualifying purchases.