Data Science Data Governance

Abstract

An increasing number of consequential decisions are made automatically by software that employs machine learning, data analytics, and artificial intelligence to discover decision rules using data. The shift to data driven systems exacerbates gaps between traditional governance and oversight processes and the realities of software-driven decision-making. And with more and more software-mediated systems turning to machine learning, data analytics, and artificial intelligence to discover decision rules using data instead of having humans code those rules by hand, this gap can exist even for the software engineers, data scientists, and system operators who design, build, deploy, and manage the machines that mediate our modern lives. Whether algorithms are approving credit applications, selecting travelers for security screening, driving a car, granting and denying visas, or determining the risk profile of an accused or convicted criminal, there is a broad societal interest in ensuring the good governance of these technologies and building accountable algorithms.