Data science of the social: How the practice is responding to ethical crisis and spreading across sectors

Thesis (Ph.D.)--University of Washington, 2018This dissertation is based on three years of ethnographic fieldwork within the Data Science Environment at the University of Washington. Employing a practice-based approach, it focuses on two processes involved in "data science of the social" that are core issues for my community of study: addressing the ethical crisis that is facing the field, and advancing data-intensive capacities across social sectors and problem spaces. In the chapters devoted to ethics, I argue that ethical sense-making in data science of the social is a form of vernacular theorizing informed by implicit understandings of sociomateriality. Recognizing data science practitioners as vernacular theorists puts academic theories of sociomateriality into conversation with practice, and allows for cross-pollination among scholars of science and technology studies, critical data scholars, and data scientists. Given pressing ethical concerns and challenges accompanying data-intensive technologies, such dialogue is both generative and necessary. The vernacular theorizing of data science practitioners yields four distinct approaches to ethics in data science of the social: data science as ethical convention, data science as ethical interrogation, data science as ethical innovation, and data science as ethical participation. I explore the conditions and processes that support two of these approaches—data science as ethical convention and data science as ethical innovation—by telling the stories of two projects from the “Data Science for Social Good” program at the University of Washington. The latter chapters of this dissertation explore how data-intensive practices and technologies are spreading to new social sectors and problem spaces. Data science practitioners are often concerned with “scaling” their work by making it replicable in new contexts. This manner of expanding the reach and scope of data science is quite distinct from the type of “scaling” explored in many studies of spatially and temporally distributed scientific work, which frequently focus on the role of information infrastructures and knowledge infrastructures in enabling large-scale scientific collaborations. But the understanding of infrastructure developed in that body of work does not sufficiently capture or explain approaches to scaling in data science of the social. Therefore, I develop the concept of exostructure as a companion to infrastructure and a key mechanism enabling scaling in data science of the social. Exostructures are made up of the components of temporary, project-based collaborations intended to spawn replication or further investment in information and knowledge infrastructures. I argue that the portable, transient, iterative, and customized nature of exostructure lends itself to processes with transformative implications for the relationships among sectors and institutions participating in their development. In particular, through exostructural arrangements, the university is playing emergent roles in the data-based knowledge society by mediating between business and government, mitigating risks for other sectors, and providing a source of intellective labor