Abstract
Data is all around us. Embodied in new knowledge production practices, changing organizational structures, and new formal institutions, the practices associated with data science, such as the collection of massive data sets, machine learning, and data mining are shaping knowledge and decisions in areas ranging from medical research to finance and national security. In response, the field of critical data studies has called for scholarship that explores the new cultural paradigm and social consequences of big data (Kitchin 2014, Beer 2016). Despite an abundance of thoughtful critiques (boyd and Crawford 2012), examinations of the discourse around data (Puschmann & Burgess, 2014), and analyses of the construction and content of algorithms (Bucher 2012), little research has investigated the on-the-ground processes and professional contexts through which data analytics and algorithms are filtered as they are transformed into decisions, actions, and consequences. As such, understanding the interplay of data analytics with epistemological contexts, professional codes and identities, and variations in organizational structure is fundamental to any attempt to articulate the relationship between data and society.
This project addresses the social consequences of data science by examining the culture and practices that unfold alongside it. Drawing on in-depth interviews with 28 data scientists, a content analysis of 33 business-to-business white papers, and a year-long ethnographic case study of medical analytics in which I observed both the developers and users of medical algorithms, I ask how data is interpreted, how data and algorithms shape decision-making in practice, and what conceptions of reality accompany the use of data analytics.
In doing so, I focus on reconstructing what I call the epistemological landscape of data science. In considering the beliefs about science, the nature of reality, the human capacity to recognize that reality, and valid ways to know the truth, I argue for a reintroduction of the analysis of subjectivities into the sociology of knowledge and studies of knowledge production. I find that the epistemological landscape and practices promoted by data scientists contain a cultural framework that elevates data science above other methods of producing knowledge and that makes it difficult to challenge the widespread data collection and analysis practices on which the data science industry thrives. However, I also find that the subjective aspects of particular professional settings greatly influence the use and interpretation of data. For example, in the case of medical analytics, I show that the epistemological orientation of evidence-based medicine in combination with a cultural value of experiential knowledge leads doctors and nurses to condition and modify their interpretations of medical analytics in ways contrary to the intentions of the developers.
This work demonstrates that the study of analytics in context is central to our understanding of how the rise and institutionalization of data-driven knowledge will affect social structures and processes. It also shows that subjective experiences, cultural orientations, and local epistemological landscapes are important aspects of how particular knowledge claims come to fruition in these specific contexts. As such, I conclude by considering how the study of epistemological landscapes can inform our attempts to reckon with life in a knowledge society. I suggest that there may not be a single manifestation of culture of the knowledge society, but a multitude that take shape in response to local cultures and pragmatic demands. Though the epistemological landscape of data science is but one manifestation, its increasing institutionalization combined with the utopian visions contained within this landscape indicate that it is poised to become a dominant mode through which meaning is constructed, claims are justified, problems are solved, and actions are chosen.
