DOCSLIB.ORG

The Pitfalls of Prediction by Greg Ridgeway

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
The Pitfalls of Prediction by Greg Ridgeway The Pitfalls of Prediction by Greg Ridgeway The criminal justice system should take advantage of the latest scientific developments to make reliable predictions. rediction is common in every- be deployed to prevent future crime, day life. We make predictions and victims wrestle with the uncer- Pabout the length of our morn- tain odds of being revictimized. ing commute, the direction of the stock market, and the outcomes There is a long history of research on of sporting events. Most of these prediction in criminology and criminal common-sense predictions rely justice, and two developments are on cognitive shortcuts — or helping the criminal justice system heuristics — that shape our expec- improve its ability to make reliable, tations of what is likely to occur in scientific predictions. First, more and the future. But these heuristics are more jurisdictions are accumulating not necessarily true; they rely on rich data and are getting better at cognition, memory and sensory linking across their data sources. impressions rather than a balanced Second, a growing set of sophis- analysis of facts. Consequently, they ticated analytic prediction tools is can result in biased predictions. available to help agencies make decisions about future events, The challenge of predicting the future unknown risks and likely outcomes. has always been at the heart of Editor's Note: This article was the criminal justice system. Judges Practitioners can now combine presented to seven law enforcement weigh the risks of releasing offend- expert assessment with data-driven agencies that were developing ers to probation, police agencies try prediction models to discern how predictive policing programs. to anticipate where officers should much risk a probationer poses, 34 NIJ JOURNAL / ISSUE NO. 271 ■ FEBRUARY 2013 determine whether a pair of illicit drug transactions signals the emer- NIJ’s Role in Predictive Policing gence of a drug market, or project whether crime will increase or decrease during the next month. More and more, police departments aw enforcement work is frequently reactive: Officers respond to are using forecasting tools as a basis Lcalls for service, control disturbances and make arrests. But law for formal predictive policing efforts; enforcement work is becoming increasingly proactive: Departments these statistical prediction methods combine data with street intelligence and crime analysis to understand inform their prevention strategies so why a problem arises and predict what might happen next if they take they can anticipate rather than react certain actions. 1 to crime. (See sidebar, “NIJ’s Role in NIJ is supporting predictive policing efforts in a number of ways: Predictive Policing,” on this page.) n Predictive policing symposiums. NIJ convened two symposiums Although the science of prediction at which researchers, practitioners and law enforcement leaders continues to improve, the work developed and discussed the concept of predictive policing and its of making predictions in criminal impact on crime and justice. Read summaries of both sessions at justice is plagued by persistent http://www.nij.gov/topics/law-enforcement/strategies/predictive- shortcomings. Some stem from policing/symposium/welcome.htm. unfamiliarity with scientific strategies n Predictive policing grants. The Chicago and Shreveport police or an over-reliance on timeworn — departments are using grants to explore data-driven policing but unreliable — prediction habits. strategies. In Phase 1, they received funding to identify a problem If prediction in criminal justice is to and develop predictive policing strategies to solve it. In Phase 2, take full advantage of the strength they were awarded additional funding to implement and evaluate of these new tools, practitioners, the strategies. For more on these grants, see http://www.nij.gov/nij/ analysts, researchers and others topics/law-enforcement/strategies/predictive-policing/symposium/ must avoid some commonplace discussion-demonstrations.htm. mistakes and pitfalls in how they make predictions. For more information: Pitfall #1: Trusting Expert n To learn more about predictive policing in general, read the NIJ Prediction Too Much Journal article “Predictive Policing: The Future of Law Enforcement?” Using data and computers to predict at http://www.nij.gov/journals/266/predictive.htm. or help experts predict shows prom- ise, but the pace of adoption has not matched that promise. Why? Perhaps we trust ourselves more than we offices. Some researchers have would vote on the 68 upcoming trust machines. found that physicians have “a high cases on the 2002 docket. Based on regard for their own decision-making their knowledge of the justices and For example, more than 30 years ability and are afraid of any competi- the ins and outs of the court, they ago, Stanford scientists developed tion from computers.”3 correctly predicted how the Supreme a pathbreaking, computer-based Court would vote on 59 percent of medical expert system that could So how do experts and machines the cases. synthesize patient features and compare in their ability to predict in therapeutic options.2 The system, the justice system? Researchers used a computer pro- called MYCIN, outperformed practi- gram to make the same prediction. tioners in selecting the right antibiotic Consider this example: A panel of The computer analyzed 628 previous treatments. Despite MYCIN’s 83 experts — law professors, deans Supreme Court cases and generated demonstrated success and similar of law schools and others who had data-derived rules.4 The researchers kinds of computer-based prediction practiced before or clerked at the created a decision-tree prediction successes, we still do not see these U.S. Supreme Court — set out to model based on a simple set of systems being used in our doctors’ predict how the U.S. Supreme Court these rules. The Pitfalls of Prediction | 35 NIJ JOURNAL / ISSUE NO. 271 ■ FEBRUARY 2013 Pitfall #2: Clinging to What You Figure 1. Decision Tree for Supreme Court Justice Learned in Statistics 101 Sandra Day O'Connor If your knowledge of prediction is Was the lower court decision liberal? limited to what gets covered in intro- ductory statistics courses, you are Yes probably unfamiliar with the predic- tion model used above. Instead, you No Reverse most likely learned how to check model assumptions and carefully test hypotheses. But when it comes Was the case from the 2nd District, to prediction, the rules are different 3rd District, D.C. or Federal Circuit? and rather simple: Are the predic- tions accurate, and can you get them Yes when you need them? You can judge No the quality of a specific prediction Affirm model by considering the following: Performance criteria. Do the Was the respondent an entity other model’s goals and constraints match than the U.S. government? the intended use? Methods that are good at predicting, for example, No whether an injury will result from a Yes mission are not necessarily the same Reverse as those that are good at predicting the number of days an officer will Was the case concerning civil rights, be out with that injury. If you are 1st Amendment, economic or federalism issue? planning a tactical unit’s staffing, it is important for you to know the No Yes expected person-hours that will be lost to injuries. Thus, using a model that can accurately predict only Affirm Reverse whether an injury will occur — and not how long an officer will be out — would be insufficient. Figure 1 shows the decision tree for So what can we take away from Accuracy. Can the model make Justice Sandra Day O’Connor. Based this example? It should lead us to accurate forecasts? More specifically, on a simple set of rules — such as question — but not necessarily the implemented model should be whether the lower court decision dismiss — the predictions of experts, better at prediction than the agency’s was liberal — the model was able to including ourselves. Of course, not current practice. For example, if cops predict how Justice O’Connor would all cases afford us the data to build are allocating resources to neighbor- decide 70 percent of the cases in predictive models. But if we have hoods where they think crime will 2002. Using similar decision trees for data that we can use to construct spike, then going forward we should the other eight justices,5 the model predictive models, then we should test whether the prediction model is correctly predicted the majority build the models and test them even better at selecting those neighbor- opinion in 75 percent of the cases, if our expert detectives, probation hoods. If a probation officer is assign- substantially outperforming the officers and others in the field indi- ing remote monitoring anklets to DUI experts’ 59 percent. The experts cate that they already know how to probationers, then we should test lost out to a machine that had a predict. They may be as surprised as whether the prediction model is few basic facts about the cases. the expert panel was in the Supreme better at picking which DUI proba- Court example. tioners will reoffend in the next six 36 | The Pitfalls of Prediction NIJ JOURNAL / ISSUE NO. 271 ■ FEBRUARY 2013 months. For a prediction model accurate in predicting, say, when and to be useful, it must outperform where a gang retaliation shooting will Prediction can play practice as usual. take place, then a more transparent a major role in model might not be worth the cost. Computation time. Can you apply This issue will be discussed further the criminal justice the prediction model in a reasonable under Pitfall #5. amount of time? Some models can system. Even small be computationally intensive to run Pitfall #3: Assuming One and use. There is little point in using Method Works Best for improvements in where a model that cannot produce predic- All Problems tions in time for them to be useful.
Load more

Recommended publications
  • Evaluation of the Shreveport Predictive Policing Experiment
    Safety and Justice Program CHILDREN AND FAMILIES The RAND Corporation is a nonprofit institution that EDUCATION AND THE ARTS helps improve policy and decisionmaking through ENERGY AND ENVIRONMENT research and analysis. HEALTH AND HEALTH CARE This electronic document was made available from INFRASTRUCTURE AND www.rand.org as a public service of the RAND TRANSPORTATION Corporation. INTERNATIONAL AFFAIRS LAW AND BUSINESS NATIONAL SECURITY Skip all front matter: Jump to Page 16 POPULATION AND AGING PUBLIC SAFETY SCIENCE AND TECHNOLOGY TERRORISM AND HOMELAND SECURITY Support RAND Purchase this document Browse Reports & Bookstore Make a charitable contribution For More Information Visit RAND at www.rand.org Explore the RAND Safety and Justice Program View document details Limited Electronic Distribution Rights This document and trademark(s) contained herein are protected by law as indicated in a notice appearing later in this work. This electronic representation of RAND intellectual property is provided for non-commercial use only. Unauthorized posting of RAND electronic documents to a non-RAND website is prohibited. RAND electronic documents are protected under copyright law. Permission is required from RAND to reproduce, or reuse in another form, any of our research documents for commercial use. For information on reprint and linking permissions, please see RAND Permissions. This report is part of the RAND Corporation research report series. RAND reports present research findings and objective analysis that ad- dress the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for re- search quality and objectivity. Evaluation of the Shreveport Predictive Policing Experiment Priscillia Hunt, Jessica Saunders, John S.
    [Show full text]
  • Predictive POLICING the Role of Crime Forecasting in Law Enforcement Operations
    Safety and Justice Program CHILDREN AND FAMILIES The RAND Corporation is a nonprofit institution that EDUCATION AND THE ARTS helps improve policy and decisionmaking through ENERGY AND ENVIRONMENT research and analysis. HEALTH AND HEALTH CARE This electronic document was made available from INFRASTRUCTURE AND www.rand.org as a public service of the RAND TRANSPORTATION Corporation. INTERNATIONAL AFFAIRS LAW AND BUSINESS NATIONAL SECURITY Skip all front matter: Jump to Page 16 POPULATION AND AGING PUBLIC SAFETY SCIENCE AND TECHNOLOGY TERRORISM AND HOMELAND SECURITY Support RAND Purchase this document Browse Reports & Bookstore Make a charitable contribution For More Information Visit RAND at www.rand.org Explore the RAND Safety and Justice Program View document details Limited Electronic Distribution Rights This document and trademark(s) contained herein are protected by law as indicated in a notice appearing later in this work. This electronic representation of RAND intellectual property is provided for non-commercial use only. Unauthorized posting of RAND electronic documents to a non-RAND website is prohibited. RAND electronic documents are protected under copyright law. Permission is required from RAND to reproduce, or reuse in another form, any of our research documents for commercial use. For information on reprint and linking permissions, please see RAND Permissions. This report is part of the RAND Corporation research report series. RAND reports present research findings and objective analysis that ad- dress the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for re- search quality and objectivity. Safety and Justice Program PREDICTIVE POLICING The Role of Crime Forecasting in Law Enforcement Operations Walter L.
    [Show full text]
  • How Will AI Shape the Future of Law?
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Helsingin yliopiston digitaalinen arkisto How Will AI Shape the Future of Law? EDITED BY RIIKKA KOULU & LAURA KONTIAINEN 2019 Acknowledgements The editors and the University of Helsinki Legal Tech Lab would like to thank the authors and interviewees for the time and effort they put into their papers. We would also like to thank the Faculty of Law at University of Helsinki and particularly the continuous kind support of Professor Kimmo Nuotio, the former dean of the faculty and the Lab’s unofficial godfather, for facilitating the Lab’s development and for helping us make the second Legal Tech Conference happen. In addition, we would like to express our gratitude to the conference sponsors, Finnish Bar Association, the Association of Finnish Lawyers, Edita Publishing, Attorneys at Law Fondia and Attorneys at Law Roschier as well as the Legal Design Summit community for their support. It takes a village to raise a conference. Therefore, we would like to thank everyone whose time and commitment has turned the conference and this publication from an idea into reality. Thank you to the attendees, speakers, volunteers and Legal Tech Lab crew members. RIIKKA KOULU & LAURA KONTIAINEN Legal Tech Lab, University of Helsinki 2019 University of Helsinki Legal Tech Lab publications © Authors and Legal Tech Lab ISBN 978-951-51-5476-7 (print) ISBN 978-951-51-5477-4 (PDF) Print Veiters Helsinki 2019 Contents Foreword 009 KIMMO NUOTIO I Digital Transformation of
    [Show full text]
  • Establishing Discriminatory Purpose in Algorithmic Risk Assessment
    NOTES BEYOND INTENT: ESTABLISHING DISCRIMINATORY PURPOSE IN ALGORITHMIC RISK ASSESSMENT Equal protection doctrine is bound up in conceptions of intent. Unintended harms from well-meaning persons are, quite simply, nonac- tionable.1 This basic predicate has erected a high bar for plaintiffs ad- vancing equal protection challenges in the criminal justice system. Notably, race-based challenges on equal protection grounds, which sub- ject the state to strict scrutiny review, are nonetheless stymied by the complexities of establishing discriminatory purpose. What’s more, the inability of courts and scholars to coalesce on a specific notion of the term has resulted in gravely inconsistent applications of the doctrine.2 States’ increasing use of algorithmic systems raises longstanding con- cerns regarding prediction in our criminal justice system — how to cat- egorize the dangerousness of an individual, the extent to which past behavior is relevant to future conduct, and the effects of racial dispari- ties.3 Integration of algorithmic systems has been defined by ambiva- lence: while they are touted for their removal of human discretion,4 they also readily promote and amplify inequalities — for example, through their consideration of protected characteristics and their interaction with existing systems tainted by legacies of inequality.5 Furthermore, algo- rithms, especially those incorporating artificial intelligence (AI), may operate in ways that are opaque, unpredictable, or not well understood. ––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 See Washington v. Davis, 426 U.S. 229, 239–40 (1976). 2 See Aziz Z. Huq, What Is Discriminatory Intent?, 103 CORNELL L. REV. 1211, 1240–63 (2018) (presenting five definitions from the literature). 3 See Sonia K.
    [Show full text]
  • CHRISTOPHER SLOBOGIN Vanderbilt University Law School, 131 21St Ave., Nashville, Tenn
    CHRISTOPHER SLOBOGIN Vanderbilt University Law School, 131 21st Ave., Nashville, Tenn. 37203-1181 Phone: (615) 343 2059; Fax: 322 6631; E-mail: [email protected] LEGAL AND PRE-LEGAL EDUCATION LL.M., University of Virginia Law School, 1979. J.D., University of Virginia Law School, 1977. A.B., Princeton University, 1973. LEGAL TEACHING EXPERIENCE Appointments: Vanderbilt University: Professor of Law, 7/08-present; Milton Underwood Professor of Law, 10/09 - present; Professor, Department of Psychiatry (secondary appointment), 2/09 - present. University of Florida: Stephen C. O’Connell Chair, 8/98-7/08; Alumni Research Scholar, 6/94-8/98; Prof., 3/87; Associate Prof., 3/85-3/87; Assistant Prof., 9/82-3/85; Affiliate Prof., Dep’t Psychiatry, 3/95-7/08. University of South Florida: Adjunct Professor, Department of Mental Health, 6/86 - present. Visitorships: Oxford University, Centre for Criminology (Fellowship), 5/17-7/17. Stanford Law School, Edwin A. Heafey Visiting Scholar, 9/06-5/07. University of California, Hastings Law School, 1/05-5/05; 1/03-5/03. University of Frankfurt Law School, Frankfurt, Germany, 4/01-7/01. University of Southern California Law School, 1/00-5/00. Monash University, Melbourne, Australia, 2/91-6/91. University of Virginia Law School, 8/88-6/89. University of Nebraska Law School, 6/88-8/88. Kiev University Law School, Kiev, Ukraine, Fulbright Scholar, 10/87-12/87. Courses Taught Criminal Procedure I & II Mental Health Law Professional Responsibility Comparative Criminal Procedure Health Law Evidence Criminal Law Law & Social Science Civil Procedure White Collar Crime Forensic Practicum Introduction to Law Teaching Awards/Evaluations: Average overall evaluation score, fall, 2003 through fall, 2020 at UF and Vanderbilt: 4.6 (out of 5) Hall-Hartman Outstanding Professor Award, Hartman Division, Vanderbilt University, 2018-2019 Chosen to create Vanderbilt-Coursera Course on “Hot Topics in Criminal Justice” (six 90-minute lectures, 5/19).
    [Show full text]
  • Implementing Predictive Policing: a Policy Analysis Approach
    Implementing Predictive Policing: A Policy Analysis Approach Item Type text; Electronic Thesis Authors Guzman, Samantha Grace Publisher The University of Arizona. Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author. Download date 28/09/2021 07:40:49 Item License http://rightsstatements.org/vocab/InC/1.0/ Link to Item http://hdl.handle.net/10150/632758 IMPLEMENTING PREDICTIVE POLICING: A POLICY ANALYSIS APPROACH By Samantha Grace Guzman ____________________ A Thesis Submitted to The Honors College In Partial Fulfillment of the Bachelors degree With Honors in Criminal Justice Studies THE UNIVERSITY OF ARIZONA M A Y 2 0 1 9 Approved by: __________________ Anne E. Boustead, JD PhD School of Government and Public Policy 1 Abstract Predictive policing techniques have grown rapidly in both sophistication and use over the past decade; yet, there is a large amount of legal and public pushback against the concept. This thesis aims to address privacy, equity and efficacy concerns by examining the accuracy and fairness of existing predictive policing mechanisms. After providing a background on both conventional and predictive policing methods, I then introduce a framework of best practices for predictive policing implementation for policy-makers and police departments to follow. To provide examples of the framework in action, I describe how various departments have implemented predictive policing and discuss the implications of such. Finally, the paper concludes with overall policy recommendations and key takeaways for future use of predictive policing technology.
    [Show full text]
  • Challenging Racist Predictive Policing Algorithms Under the Equal Protection Clause
    41254-nyu_94-3 Sheet No. 117 Side B 05/14/2019 08:58:42 \\jciprod01\productn\N\NYU\94-3\NYU306.txt unknown Seq: 1 13-MAY-19 13:23 CHALLENGING RACIST PREDICTIVE POLICING ALGORITHMS UNDER THE EQUAL PROTECTION CLAUSE RENATA M. O’DONNELL* Algorithms are capable of racism, just as humans are capable of racism. This is particularly true of an algorithm used in the context of the racially biased criminal justice system. Predictive policing algorithms are trained on data that is heavily infected with racism because that data is generated by human beings. Predictive policing algorithms are coded to delineate patterns in massive data sets and subse- quently dictate who or where to police. Because of the realities of America’s crim- inal justice system, a salient pattern emerges from the racially skewed data: Race is associated with criminality in the United States. Because of the “black-box” nature of machine learning, a police officer could naively presume that an algorithm’s results are neutral, when they are, in fact, infected with racial bias. In this way, a machine learning algorithm is capable of perpetuating racist policing in the United States. An algorithm can exacerbate racist policing because of positive feedback loops, wherein the algorithm learns that it was “correct” in associating race and criminality and will rely more heavily on this association in its subsequent iterations. This Note is the first piece to argue that machine learning-based predictive policing algorithms are a facial, race-based violation of the Equal Protection Clause. There will be major hurdles for litigants seeking to bring an equal protection challenge to these algorithms, including attributing algorithmic decisions to a state actor and overcoming the proprietary protections surrounding these algorithms.
    [Show full text]
  • Predictive Policing and Reasonable Suspicion
    Emory Law Journal Volume 62 Issue 2 2012 Predictive Policing and Reasonable Suspicion Andrew Guthrie Ferguson Follow this and additional works at: https://scholarlycommons.law.emory.edu/elj Recommended Citation Andrew G. Ferguson, Predictive Policing and Reasonable Suspicion, 62 Emory L. J. 259 (2012). Available at: https://scholarlycommons.law.emory.edu/elj/vol62/iss2/1 This Article is brought to you for free and open access by the Journals at Emory Law Scholarly Commons. It has been accepted for inclusion in Emory Law Journal by an authorized editor of Emory Law Scholarly Commons. For more information, please contact [email protected]. FERGUSON GALLEYSPROOFS2 2/19/2013 9:49 AM PREDICTIVE POLICING AND REASONABLE SUSPICION ∗ Andrew Guthrie Ferguson INTRODUCTION .............................................................................................. 261 I. PREDICTIVE POLICING: AN INTRODUCTION ........................................ 265 A. Predictive Policing: In Context .................................................. 270 1. Intelligence-Led Policing and Theories of Crime and Place ..................................................................................... 271 2. Predictive Models of Crime .................................................. 276 a. Near Repeat Theory ....................................................... 277 b. Risk Terrain Modeling ................................................... 281 B. Predictive Policing: Future Cases ............................................. 284 II. PREDICTION
    [Show full text]
  • Predictive Policing: Preventing Crime with Data and Analytics, (Washington, DC: IBM Center for the Business of in Governmentmanagement Government, 2013)
    Forum: Six Trends Driving Change This article is adapted from Jennifer Bachner, Predictive Policing: Preventing Crime with Data and Analytics, (Washington, DC: IBM Center for The Business of in GovernmentManagement Government, 2013). Predictive Policing: Preventing Crime with Data and Analytics By Jennifer Bachner The history of quantitative crime analysis spans centuries. prevented police departments from integrating this tool into Crime mapping first appeared in the 19th century. In 1829, everyday police work until recently. an Italian geographer and French statistician designed the first maps that visualized crime data. The maps included three Today, the availability of massive data sets, data storage, years of property crime data as well as education information sophisticated software, and personnel that can both perform obtained from France’s census. The maps revealed a positive analyses and communicate actionable recommendations correlation between these two layers of information; areas to officers in the field has rendered crime analysis a central with higher levels of education experienced a higher inci- component of modern policing. Further, collaborative efforts dence of property crimes. between police officers, scholars, and businesses have led to the development of analytical techniques that have strong The discipline of crime analysis emerged following the theoretical foundations; accompanying tools, such as soft- formation of London’s Metropolitan Police, the first organized ware programs, enable their widespread use. law enforcement service. The service’s detective branch, formed in 1842, was tasked with using pattern recognition to The Role of Predictive Analytics in Crime prevent and solve crimes. Formal police departments were Prevention established throughout the U.S. in the 1850s, though their use of analytical techniques lagged behind London’s.
    [Show full text]
  • Disparate Impact in Big Data Policing
    DISPARATE IMPACT IN BIG DATA POLICING Andrew D. Selbst* Data-driven decision systems are taking over. No institution in society seems immune from the enthusiasm that automated decision-making generates, including—and perhaps especially—the police. Police departments are increasingly deploying data mining techniques to predict, prevent, and investigate crime. But all data mining systems have the potential for adverse impacts on vulnerable communities, and predictive policing is no different. Determining individuals’ threat levels by reference to commercial and social data can improperly link dark skin to higher threat levels or to greater suspicion of having committed a particular crime. Crime mapping based on historical data can lead to more arrests for nuisance crimes in neighborhoods primarily populated by people of color. These effects are an artifact of the technology itself, and will likely occur even assuming good faith on the part of the police departments using it. Meanwhile, predictive policing is sold in part as a “neutral” method to counteract unconscious biases when it is not simply * Postdoctoral Scholar, Data & Society Research Institute; Visiting Fellow, Yale Information Society Project. J.D. 2011, University of Michigan. For helpful comments and insights on earlier drafts, I would like to thank Micah Altman, Jack Balkin, Jane Bambauer, Solon Barocas, Rabia Belt, Kiel Brennan-Marquez, William Buzbee, Danielle Citron, Julie Cohen, Jessica Eaglin, Gautam Hans, Don Herzog, Mark MacCarthy, Deirdre Mulligan, Leonard Niehoff, Paul Ohm, Christopher Slobogin, Katherine Strandburg, David Super, Michael Veale, David Vladeck, Robin West, and the participants at the Georgetown Fellows Colloquium, Georgetown Summer Faculty Workshop, Michigan Future Law Professors Workshop, and 2017 Privacy Law Scholars’ Conference.
    [Show full text]
  • Dirty Data, Bad Predictions: How Civil Rights Violations Impact Police Data, Predictive Policing Systems, and Justice
    RICHARDSON_ETAL-PROOF.DOCX(DO NOT DELETE) 4/22/2019 2:07 PM DIRTY DATA, BAD PREDICTIONS: HOW CIVIL RIGHTS VIOLATIONS IMPACT POLICE DATA, PREDICTIVE POLICING SYSTEMS, AND JUSTICE RASHIDA RICHARDSON,* JASON M. SCHULTZ† & KATE CRAWFORD‡ Law enforcement agencies are increasingly using predictive policing systems to forecast criminal activity and allocate police resources. Yet in numerous jurisdictions, these systems are built on data produced during documented periods of flawed, racially biased, and sometimes unlawful practices and policies (“dirty policing”). These policing practices and policies shape the environment and the methodology by which data is created, which raises the risk of creating inaccurate, skewed, or systemically biased data (“dirty data”). If predictive policing systems are informed by such data, they cannot escape the legacies of the unlawful or biased policing practices that they are built on. Nor do current claims by predictive policing vendors provide sufficient assurances that their systems adequately mitigate or segregate this data. In our research, we analyze thirteen jurisdictions that have used or developed predictive policing tools while under government commission investigations or federal court monitored settlements, consent decrees, or memoranda of agreement stemming from corrupt, racially biased, or otherwise illegal policing practices. In particular, we examine the link between unlawful and biased police practices and the data available to train or implement these systems. We highlight three case studies: (1) Chicago, an example of where dirty data was ingested directly into the city’s predictive system; (2) New Orleans, an example where the extensive evidence of dirty policing practices and recent litigation suggests an extremely high risk that dirty data was or could be used in predictive policing; and (3) Maricopa County, where despite extensive evidence of dirty policing practices, a lack of public transparency about the details of various predictive policing systems restricts a proper assessment of the risks.
    [Show full text]
  • Collecting, Analyzing, and Responding to Stop Data: a Guidebook for Law Enforcement Agencies, Government, and Communities
    Collecting, Analyzing, and Responding to Stop Data: A Guidebook for Law Enforcement Agencies, Government, and Communities Marie Pryor, PhD Farhang Heydari, JD Center for Policing Equity Policing Project at New York University School of Law Philip Atiba Goff, PhD Barry Friedman, JD Center for Policing Equity Policing Project at New York University School of Law Contributors to this Guidebook Center for Policing Equity Christopher Mebius, Lee Dobrowolski, Tracie Keesee, R. Nicole Johnson-Ahorlu, John Tindel, Dominique Johnson Policing Project at New York University School of Law Brian Chen, Carrie Eidson, Ariele LeGrand, Maria Ponomarenko, Christina Socci, Robin Tholin California Department of Justice Randie Chance, Jenny Reich, Kevin Walker, Trent Simmons, Amanda Burke This project was supported by cooperative agreement number 2016-CK-WX-KOJ6, awarded by the Office of Community Oriented Policing Services, U.S. Department of Justice. The opinions contained herein are those of the authors and do not necessarily represent the official positions or policies of the U.S. Department of Justice. References to specific agencies, companies, products, or services should not be considered an endorsement by the authors or the U.S. Department of Justice. Rather, the references are illustrations to supplement discussion of the issues. The internet references cited in this publication are valid as of the date of this publication. Given that URLs and websites are in constant flux, neither the authors nor the Office of Community Oriented Policing Services can vouch for their current validity. Copyright © 2020 Center for Policing Equity and Policing Project at New York University School of Law. The U.S.
    [Show full text]