Thinking About Data-Driven Policing in an Era of Social Justice
Over the past several years, police in the United States have come under increased scrutiny. The Black Lives Matter movement and those sympathetic to its cause have charged police with bias against black and brown people. Community leaders, scholars, and activists assert that American policing practices overpolice black and brown neighborhoods, and are overly aggressive towards black and brown people.
The high-profile deaths of Trayvon Martin, Eric Garner, Sandra Bland, and other people of color have turned what used to be a concern only within black and brown communities into a national problem. In 2020, George Floyd and Breonna Taylor’s deaths raised awareness to new heights. Some activists began calling for a defunding of police departments.
One possible way for police to address concerns of bias is through data-driven policing. By data-driven policing, I am referring to police departments making decisions based upon the collecting of data and using computers to find patterns in that data. These patterns then inform decisions on where to place police officers or who to investigate. The hope is that data driven decisions, based on objective information, will reduce the amount of bias in policing.
In this essay, I discuss scholarship about data-driven policing from academics. These articles are written by legal scholars, data scientists, and criminologists. I end by providing some possible ways in which data-driven policing can become a productive part of law enforcement practices.
What is Data-Driven Policing?
An example of data-driven policing would be using police incident reports from 2020 to understand where the most crimes were committed and then placing more patrol cars in those neighborhoods in 2021. A sample of an incident report, from the Bakersfield, California police department, is shown below. The officer who responds to an incident will record the details on a form like the one shown. Each cell on the form will eventually be transferred into a database with other incident reports. One approach would be to look at the time and location of crime during the month of June in 2020 and then position manpower at those same times and locations in June of 2021. The logic being that the same external factors that generated crime in June 2020 at those times and places, will reoccur in 2021. For this reason, data-driven policing is often called predictive policing.
There are more innovative ways of using data. For example, suppose a police department is attempting to identify the leaders of a local gang. Suppose they have arrested or surveilled several gang members, and through either interrogation, social media accounts, or personal observation, they have a list of their friends, family, and associates. If they see that a person is connected to many gang members, this gives police a clue that they are important and maybe a leader. Police have always done this. But now with computer analyses, they can build more precise, statistically sound social network models.
What Do We Know About Data-Driven Policing
Some scholars argue that the use of data-driven policing does not remove bias but instead entrenches it. For example, Isaac and Dixon (2019) argue that big data analysis tools “reinforce, rather than reimagine, existing police practices. Their expanded use could lead to further targeting of communities or people of color.” They give the example of the city of Chicago. Chicago has employed software that places police in high crime areas and identifies people heading down a criminal path. They argue that the data used to predict high crime areas or potential criminals reflects past over-policing more than any real criminal activity. And so, because racist police practices overpoliced black and brown neighborhoods in the past, this appears to mean these are high crime areas, and even more police are placed there. They argue that a way forward is to make the algorithms transparent so that the algorithms’ users can evaluate it for bias.
Data driven, predictive policing is not only an issue in the United States. Countries all over the world are using data, computers, and algorithms to anticipate where crime is likely to happen. United Kingdom professor Mike Rowe draws similar conclusions about bias in algorithms. Rowe discusses the bias that was baked into the U.K. police department’s data to determine if a suspect should be referred to a rehabilitation program instead of being kept in custody. If a suspect lived in a low socioeconomic zip-code, they were less likely to be referred to a rehab program. This, however, reproduces inequalities because being born into a low-income area is a bias against poor people (who are disproportionately people of color). An interesting analogy that Rowe provides is to compare big data policing with detective work. Rowe points out that detectives have done the same things as algorithms — taking in information from a variety of sources and finding patterns. Big data and algorithms simply increase the power and magnitude of this information processing.
Legal scholar Andrew Ferguson (2020), the author of The Rise of Big Data Policing, also believes that big data in policing negatively impacts marginalized groups. Big data tools, he argues, have “exacerbated rather than reduced bias, overreach, and abuse in policing, and they pose a growing threat to civil liberties.” The rationale for using data-driven policing is a good one, he suggests, as it is grounded in attempt to be cost-efficient and address racism in policing. However, he warns that as technologies become more sophisticated, citizen’s right to privacy is endangered. Because these technologies are disproportionately directed at communities of color, black and brown people are most at risk of losing civil liberties.
Other scholars share the worry about the loss of civil liberties. Data scientist H.V. Jagadish Bernard (2019) takes a middle ground between those who support data-driven policing and those who do not. He argues that predictive policing does indeed reduce crime. However, at what cost? He writes:
“Suppose there’s a one-in-a-million chance that a typical citizen will commit a murder, but there is a one-in-a-thousand chance that Tyrone will. That makes him a thousand times as likely to commit a murder as a typical citizen. So it makes sense statistically for the police to focus their attention on him. But don’t forget that there is only a one-in-a-thousand chance that he commits a murder. For a thousand such “suspect” Tyrones, there is only one who is a murderer and 999 who are innocent. How much are we willing to inconvenience or harm the 999 to stop the one?”
Bernard’s focus on probability helps us see how data-driven policing is both useful — we are statistically more accurate when using it, and harmful — we run a great risk of discriminating against people, especially people of color.
Criminal justice professors David Pyrooz and James Densley (2020) also take this middle ground. Their concern is that the push to eliminate gang databases on charges of being racist will decrease police effectiveness to deal with crime. Pyrooz and Densley point out that gang members commit a disproportionate number of crimes. Moreover, 13% of all homicides in the United States are gang-related. And in New York City 2017, 50% of all shootings were gang-related. Given this, police must have access to data on gangs.
They argue that the best way forward is not to eliminate databases but to adopt a more accurate and systematic data collection process. “The opposite of bad data is not no data, but good data”, they argue.
Data-driven policing can make policing more efficient and, at least theoretically, less biased. Since the early 2000’s law enforcement agencies globally have been embracing some type of data-driven tool. Scholars critical of this push question the objectivity of predictive policing. Some argue that the data is biased and therefore the predictions are biased. There is also a concern that using big data tools is a threat to civil liberties, as unwarranted surveillance violates privacy rights.
This does not mean, however, that there is no place for data-driven policing. It means that we need to be more critical of the entire process — from how we collect data, to how we analyze it, to thinking about its moral and ethical consequences. Professors David Pyrooz and James Densley argue that “The opposite of bad data is not no data, but good data.”
I believe the opposite of bad data-driven policing is not no data-driven policing but good data-driven policing.
One way to move toward good data-driven policing is with transparency and constant human intervention. The process by which we collect data and the algorithms used to produce conclusions must be as open to public scrutiny as possible. Moreover, decisions must always be filtered through a human, moral actor. The analogy that Dr. Rowe uses, arguing that detectives are information processors, is instructive here. It suggests to me that instead of using data-driven policing to give us the final decision on where we need to police it can give us the final suggestion. Then, individual law enforcement officers trained properly to understand the historical and cultural context within which they police, can use this suggestion to make the best decision.
[All References Are From “The Conversation”]
Andrew Guthrie Ferguson Professor of Law. (2020, July 01). High-tech surveillance amplifies police bias and overreach. Retrieved January 23, 2021, from https://theconversation.com/high-tech-surveillance-amplifies-police-bias-and-overreach-140225
David Pyrooz Assistant Professor, & James Densley Associate Professor of Criminal Justice. (2020, April 01). Is gang activity on the rise? A movement to abolish gang databases makes it hard to tell. Retrieved January 23, 2021, from https://theconversation.com/is-gang-activity-on-the-rise-a-movement-to-abolish-gang-databases-makes-it-hard-to-tell-99252
H.V. Jagadish Bernard A Galler Collegiate Professor of Electrical Engineering and Computer Science. (2019, May 21). The promise and perils of predictive policing based on big data. Retrieved January 23, 2021, from https://theconversation.com/the-promise-and-perils-of-predictive-policing-based-on-big-data-48366
Mike Rowe Professor of Criminology. (2020, July 07). A.I. profiling: The social and moral hazards of ‘predictive’ policing. Retrieved January 23, 2021, from https://theconversation.com/ai-profiling-the-social-and-moral-hazards-of-predictive-policing-92960
William Isaac Ph.D. Candidate in Political Science, & Andi Dixon Ph.D. Student in Communications. (2019, August 27). Why big-data analysis of police activity is inherently biased. Retrieved January 23, 2021, from https://theconversation.com/why-big-data-analysis-of-police-activity-is-inherently-biased-72640