Presented By O'Reilly and Cloudera
Make Data Work
September 26–27, 2016: Training
September 27–29, 2016: Tutorials & Conference
New York, NY

Thinking outside the black box: The imperative for accountability and transparency in predictive analytics

Brett Goldstein (University of Chicago)
2:55pm–3:35pm Wednesday, 09/28/2016
Law, ethics, governance
Location: River Pavilion Level: Non-technical
Average rating: *****
(5.00, 3 ratings)

Prerequisite knowledge

  • A general familiarity with how predictive analytics are being applied in the public sector
  • What you'll learn

  • Understand the important challenges to bringing greater transparency to data science
  • Learn how the market for predictive tools is changing and the opportunity for open source models is moving forward
  • Description

    With rapid advances in the field of data science and the availability of real-time streaming data, the specter of a data-driven dystopia looms larger than ever. Mainstream media, civil rights advocates, and watchdog groups of all political persuasions are increasingly questioning the legitimacy of proprietary predictive tools that are widely used in areas from law enforcement to healthcare.

    While some of these concerns are overly sensationalized, many are grounded in legitimate worry about embedded bias and lack of transparency. The market for black box tools is dying fast. The public is demanding greater insight into how and why decisions are made, creating an environment where the use of such software is rapidly becoming politically untenable, particularly in law enforcement. At the same time, competitive markets for predictive tools should encourage transparency into the performance of the underlying analytics in order to ensure optimal performance. So, how can we usher in a future of data-driven decision making that is characterized by more—not less—accountability and accessibility?

    Brett Goldstein discusses the imperative to couple new developments in data science with a renewed commitment to transparency and open source. In particular, Brett focuses on the example of CrimeScape, an open source deployment tool that optimizes policing resources transparently in real time. In contrast to existing crime analytics software, CrimeScape’s data, model, predictions, and accuracy are published online.

    Brett explores the benefits and challenges of bringing a sustainable open source approach to predictive analytics, covering issues such as the security of sensitive data, how to develop transparent model performance metrics and evaluation, and how to balance privacy with transparency within an open source framework. Brett concludes with a discussion of how these lessons can be applied more broadly in other contexts beyond crime prediction.

    Photo of Brett Goldstein

    Brett Goldstein

    University of Chicago

    Brett Goldstein is a leader in enterprise architecture, big data analytics, and government technology with 15 years of experience in operations, management, and leadership in technical environments in both the public and private sector. Brett was recently named the inaugural recipient of the Fellowship in Urban Science at the University of Chicago’s Harris School of Public Policy. As a senior fellow in urban science, he will focus on issues of computation and public policy to inform better decision making in government. Previously, Brett was the commissioner and chief information officer of the Chicago Department of Innovation and Technology (DoIT), appointed by Mayor Rahm Emanuel to accelerate Chicago’s growth as a global hub of innovation and technology. During his tenure as Chicago’s CIO, Brett successfully worked toward a comprehensive consolidation of technology while rapidly accelerating the role of innovation in government. His achievements have included changing Chicago’s technology strategy to include cloud environments, and reshaping the IT portfolio to include advanced analytics with a focus on urban prediction. Brett was also the chief data officer for the City of Chicago, the first position of this kind for a major municipality, where he led the city’s data strategy to help improve the way the city’s information works for its residents.

    Before coming to City Hall, Brett was one of the youngest commanders in the Chicago Police Department, where he founded and directed the department’s Predictive Analytics Group, which aimed to predict violent crime patterns. Previously, Brett was an early employee with OpenTable, where he played an integral role in scaling the operation from a handful of restaurants in San Francisco to a network that operates worldwide. He holds a bachelor’s degree from Connecticut College, an MS in criminal justice from Suffolk University, and an MS in computer science from University of Chicago. Brett is pursuing his PhD in criminology, law, and justice at the University of Illinois-Chicago. He resides in Chicago with his wife and three children.