How can government better use data to produce evidence, while protecting privacy for individuals who are subjects of those data? This question, tackled last year by the U.S. Commission on Evidence-Based Policymaking, re-emerged during a hearing in the House Education and the Workforce Committee on January 30th. During the hearing the committee examined government’s existing privacy protections and received testimony from a panel about strengths and potential challenges regarding education data.
Privacy protections have long been viewed by the public and government officials as an important feature of the evidence-building enterprise. With enactment of the Privacy Act of 1974 and the Confidential Information Protection and Statistical Efficiency Act of 2002, among other laws, the federal government has instituted clear procedures for protecting the confidentiality of individuals and businesses.
Among the witnesses the committee heard from Wednesday was former commission member Paul Ohm, a Georgetown law professor. Ohm was appointed by President Obama to the commission as one of the five required privacy experts and invited as a Republican witness.
Privacy protections have long been viewed by the public and government officials as an important feature of the evidence-building enterprise.
Ohm’s message to the committee reflected the bipartisan and unanimous recommendations the commission provided to Congress and President Trump in September 2017, which concluded that while existing protections are strong they are far from perfect. The recommendations the commission offered sought to encourage greater use of the data government already collects, in conjunction with stronger privacy and legal protections, increased accountability, and enhanced transparency.
A Balance Between Evidence Building and Privacy
Ohm’s testimony this week renewed emphasis on the commission’s conclusions that constant attention is needed to reviewing and managing privacy risks posed by the use of confidential government data for evidence-based policymaking.
“The hard work required of evidence-based policymaking requires us to grant researchers access to data about the private activities of the citizens and residents of this country,” Ohm said in his written statement. “These data can sometimes—not always—reveal the sensitive activities, habits, relationships, and even thoughts and aspirations of individuals, the kind of data that can lead to great harm if allowed to disseminate too far.”
Ohm describes how the approaches that enable analysis of confidential data to achieve a public good can also cause harm without sufficient restraints. The commission recommendations issued last September similarly highlight a theme of responsible use and procedures that introduce “friction” in decision-making processes for what data are made available for research.
Take, for example, the commission’s recommendation to establish a National Secure Data Service (see recommendation 2-1). The commission intended the Service to bring data together for approved projects conducted by qualified researchers for as little time as possible, and without providing researchers access to confidential information about individuals in the data files. The process for implementing the Service outlined numerous features to limit access to data and to reduce potential privacy risks.
Similarly, the commission specifically recommended against establishing a government-wide clearinghouse of administrative and survey data collected by government, recognizing the potential threats to privacy such a clearinghouse could pose in the future (see recommendation 2-2).
Applications for Privacy Protective Technologies
New technologies are emerging that may increase the ability of government to analyze more confidential data while simultaneously strengthening privacy protections. Among the approaches cited by Ohm that hold “great promise” for improving how government protects data, is secure multi-party computation (SMC).
SMC is an approach that enables statistical analysis of virtually linked data files maintained separately at multiple sites, without exposing any sensitive information from within those files. The technique relies on cryptography which enables the generation of statistics using data controlled by different data owners without sharing the data with each other. While the mathematical and computational approach has been around for decades, there are few examples of successful deployment of the technology within government to date.
The commission specifically called on government to better support state-of-the art technologies that can simultaneously enable more data analysis and stronger privacy protections, including stronger cybersecurity and confidentiality (see recommendation 3-2). The commission further suggested that the new Data Service could lead both the implementation and application of these types of approaches across government.
Ohm’s testimony was echoed by Marc Rotenberg and Christine Bannan from the Electronic Privacy Information Center, who previously endorsed the privacy recommendations of the commission, stating that “supporting adoption of [privacy-enhancing technologies] are key to protecting personal information.”
Government has a long way to go before deployment of approaches like SMC can be widespread, but pilots and small-scale demonstrations can be a productive first step to proving the relevance of the technology within the government context. The potential benefits SMC holds for both research and privacy justify the near-term effort.