The stakes of defining data science

• In Event: Critical Data Studies: Ethics and Human Contexts of Data Science I

Abstract

The term “data science” has exploded in popularity, but there is still substantial ambiguity about what data science ‘really’ is -- it means anything at all. Originally envisioned by funders and key faculty to be broadly inclusive “trading zones” (Kellogg et al,. 2006) in which data science techniques could bring disparate fields together, we explore how the work of defining data science both adds and removes collaborative friction. Moving from a single conceptual definition of data science into the lived reality of developing curricula, securing funding, building infrastructure, charting career paths, and working out interdisciplinary collaboration reveals what is at stake in the definition of key terms.

Like any discipline or profession, boundary work is done to define and delimit the contours of data science (Gieryn, 1983). Drawing on ethnographic fieldnotes and qualitative interviews at three universities with data science initiatives, we analyze debates about definitions of data science and the role of data scientists across contexts. These definitional issues reveal underlying intellectual, scientific, and cultural differences between disciplines. We walk through a series of five vignettes that capture the recent evolutionary history of data science, describing what is at stake for those employing and impacted by applications of these methods.

We argue that the way these disciplines have addressed their internal challenges around defining data science has shaped the constituencies and projects that can be incubated in the data science environments.

Authors

• Laura Noren, Obsidian Security

• R. Stuart Geiger, UC-Berkeley Institute for Data Science