Predict and surveil: Data, discretion, and the future of policing. By BrayneS, New York, NY: Oxford University Press. 2020. pp. 288. $29.95 (hard cover). ISBN : 9780190684099
基于对洛杉矶警局的五年田野调查，揭示大数据分析在警务中的主观性和偏见，指出其未能减少不平等，并提出改进数据使用的六项建议，对公共管理者有借鉴意义。
Policing systems “increasingly rely on big data and automated systems to decide who, when, and where to police” (p. 2). Software programs and algorithms that leverage big data are often marketed and perceived as inherently objective tools for decision-making and a cure-all for structural biases embedded in public systems. However, Brayne dispels this notion by showing the social, subjective nature by which policing data are collected, stored, analyzed, and deployed in practice. She takes a clear stance that big data analytics has yet to deliver on its promise of reducing inequities in the policing system, based on evidence from her five-year ethnography of the Los Angeles Police Department (LAPD) and a synthesis of the existing literature. Brayne conducted interviews, ride-alongs, observations, and archival research with the LAPD, and also spent time off-duty with police officers, interviewed federal and industry experts in policing and surveillance, and examined the legal frameworks surrounding big data and policing. This constellation of sources allows Brayne to paint a nuanced picture of the micro and macro forces influencing how big data analytics are being (mis)used in policing. In addition, she articulates potential solutions for this field to course correct that are relevant for any public administrator to consider, given the proliferation of big data analytics across government systems. To build up to high-level solutions, Brayne first traces the history of data collection and use in policing, paying particular attention to the increasing reliance on private firms to provide software programs, algorithmic tools, and big datasets. This private industry creep has allowed the LAPD to extend surveillance beyond just those under suspicion for criminal activity (i.e., dragnet surveillance) and to use people- and place- based algorithms to predict crime (i.e., directed surveillance). Dragnet and directed surveillance allow police to retrace one's digital footprint and forecast potential criminal hot spots at an increasingly granular level, which is highly valuable for solving crimes. However, these surveillance strategies are also problematic for three key reasons: (1) the development and ownership of surveillance tools within the private sphere allows for a lack of transparency and oversight that is not normally permissible in public agency operations, (2) the underlying data these tools rely on contain well-documented biases (e.g., police records) and/or are being reused from other unrelated systems without a clear purpose (e.g., bankruptcy records, vehicle registrations, personal property records, name and address combinations), and (3) citizens do not have the opportunity to explicitly consent to or refuse their non-policing data ending up in police databases. Given that private industry creep affects government systems broadly, Brayne's observations of problematic data use within the LAPD can be translated into relevant questions for all public administrators to reflect upon: To what extent are private contractors and providers being held to government standards for transparency and oversight? What biases are baked into data used for decision-making, from the time data is collected and analyzed to the ways in which data-driven responses are implemented? How can citizens weigh in on an agency or program's data use, and how are they being protected (or not) from harmful uses of data through governance and legal frameworks? This leads to a key strength of the book as a whole—Brayne's ability to use the specific case of the LAPD to craft a broadly applicable narrative about data use in public systems. She offers “six provocations meant to send readers into the world ready to seek change,” such as “slow down” and “use data to direct nonpunitive interventions” (p. 141–145). These are pertinent points to wrestle with across the public sector and among researchers, evaluators, and analysts who rely on government-held data. Within these provocations Brayne does not encourage public systems to wholly reject big data analytics, but rather, to deliberately assess the biases inherent to different datasets and the potential unintended consequences before deploying tools. Moreover, public systems should start with a clear purpose for data use and craft the operationalization of data around that purpose. For example, reducing crime, cutting costs, and ensuring equitable outcomes are distinct goals that necessitate distinct and sometimes competing strategies. Brayne is not ambivalent about the types of goals police departments should pursue, though. She advocates that data are used to target programs and services that help people rather than penalize them. When applying this to a broader public administration context, this could mean using data to evaluate which neighborhoods most need a new childcare facility, bus stop, behavioral health center, or other form of community investment, instead of using data to increase surveillance, monitoring, risk score generation, and other punitive measures. Another strength of the book is that Brayne's ethnographic methods provide deep insight into the institutional side of policing. This perspective is hard to capture as robustly as she was able, given that institutions are often reluctant to invite increased scrutiny and can more easily shield themselves from research compared with the people that interact with their systems. However, Brayne's approach necessarily favors some voices over others. For instance, community advocates and data activists are not represented in this work, and neither are the voices of citizens served by the LAPD. Seeking out these perspectives is beyond the scope of the book but is still imperative for police departments and other public systems to do when developing data use strategies. Additionally, this book does not provide the detailed, practical guidance that public administration audiences will need in order to turn Brayne's aspirational provocations into on-the-ground practices and policies. The reality is that changing institutionalized data collection, analysis, and culture is difficult and specific to the local context; and learning about the agency's underlying data landscape is only the first of many steps toward improving it. This point is not readily acknowledged by Brayne, which may lead general audiences to walk away from the book without a full appreciation for what it will take for public systems to meaningfully change. That is not to negate the imperativeness of Brayne's implications for police departments and other government agencies but to underscore the need for additional resources, advocacy, research, and leadership on making data use more equitable. Predict and Surveil: Data, Discretion, and the Future of Policing illustrates the growing entanglement between private industry and policing, pinpoints the problematic uses of data and surveillance tools, and offers a balanced approach to reducing harms caused by data use. Authors covering this subject may find it easier to veer towards extremes, identifying “data-driven” policing as either a panacea or an irreparably flawed technique” (p. 117). However, Brayne eloquently articulates a clear stance about what is problematic within the policing system while also offering potential solutions. “The reality is that the implications for inequality ultimately depend on recognizing bias in data, privileging transparency and community engagement, solving for the right problems, implementing correct interventions, and constantly reevaluating our data systems” (p. 117). None of this is easy, but Brayne demonstrates why it is highly necessary. Sharon Zanti is a social welfare doctoral student at the University of Pennsylvania's School of Social Policy and Practice. She is also a doctoral fellow at Actionable Intelligence for Social Policy, an initiative that supports state and local governments in cross-sector data sharing and integration to improve policymaking.