Sharing Private Data for Public Good
Data collaboratives,” an emerging form of partnership in which participants exchange data for the public good, have huge potential to benefit society and improve artificial intelligence. But they must be designed responsibly and take data-privacy concerns into account.
NEW YORK – After Hurricane Katrina struck New Orleans in 2005, the direct-mail marketing company Valassis shared its database with emergency agencies and volunteers to help improve aid delivery. In Santiago, Chile, analysts from Universidad del Desarrollo, ISI Foundation, UNICEF, and the GovLab collaborated with Telefónica, the city’s largest mobile operator, to study gender-based mobility patterns in order to design a more equitable transportation policy. And as part of the Yale University Open Data Access project, health-care companies Johnson & Johnson, Medtronic, and SI-BONE give researchers access to previously walled-off data from 333 clinical trials, opening the door to possible new innovations in medicine.
These are just three examples of “data collaboratives,” an emerging form of partnership in which participants exchange data for the public good. Such tie-ups typically involve public bodies using data from corporations and other private-sector entities to benefit society. But data collaboratives can help companies, too – pharmaceutical firms share data on biomarkers to accelerate their own drug-research efforts, for example. Data-sharing initiatives also have huge potential to improve artificial intelligence (AI). But they must be designed responsibly and take data-privacy concerns into account.
Understanding the societal and business case for data collaboratives, as well as the forms they can take, is critical to gaining a deeper appreciation the potential and limitations of such ventures. The GovLab has identified over 150 data collaboratives spanning continents and sectors; they include companies such as Air France, Zillow, and Facebook. Our research suggests that such partnerships can create value in three main ways.
For starters, data collaboratives can improve situational and causal analysis. Their unique collections of data help government officials better understand issues such as traffic problems or financial inequality, and design more agile and focused evidence-based policies to address them.
Moreover, such data exchanges enhance decision-makers’ predictive capacity. Today’s vast stores of public and private data can yield powerful insights into future developments and thus help policymakers plan and implement more effective measures.
Finally, and most important, data collaboratives can make AI more robust, accurate, and responsive. Although analysts suggest AI will be at the center of twenty-first-century governance, its output is only as good as the underlying models. And the sophistication and accuracy of the models generally depend on the quality, depth, complexity, and diversity of data underpinning them. Data collaboratives can thus play a vital role in building better AI models by breaking down silos and aggregating data from new and alternative sources.