“Your Spectrogram Looks Like My Genomic Sequence Data” and Other Moments from the Trading Zones of Data Science

• In Event: The Agenda(s) of Open Science

Abstract

Data science is often defined using images such as Drew Conway’s venn diagram (2013) to signal the kind of work that draws on multiple sets of skills and knowledge traditions. Yet it is simultaneously often framed around the promise of developing a coherent set of generalizable tools and methodologies that can be applied across many fields. How can we characterize the emergent interdisciplinary spaces of data science and what are the key practices of interdisciplinarity in data science? This work is based on an ethnographic study of data science practice and culture in academia, as part of a broader funded initiative to develop data science environments (Moore/Sloan Data Science Environments). For participants in and funders of the data science environments, interdisciplinarity and collaboration are core defining features of what it is to do and advance data science. We draw on Collins, Evans, and Gorman’s trading zone model to identify the locations of interdisciplinarity, or in their words, the trading zones “in which communities with a deep problem of communication manage to communicate.” (2010, p. 8). We find evidence of all four kinds of trading zones proposed in the model across the data science environment, including subversive, fractionated, interlanguage, and enforced (Collins, Evans, & Gorman, 2010). We build from these trading zones to situate the interdisciplinary practices we observe within a broader network of power relationships and culture.


Drawing on over two years of participant observation within the data science environments and over 100 interviews with affiliated researchers, we identify key practices of interdisciplinarity that characterize the emergence of data science across our fieldsite. To be sure, these observed practices were not always successful instances of interdisciplinarity. They represent the varied ways in which researchers experience interdisciplinarity in data science. The first is the development of a pidgin language (Galison 1997) around and through data that allows for idea exchange and knowledge production to occur via data both as a result of and in spite of differences in language and culture. For example, for many researchers, a language of data science made the problems across many different fields visible and thinkable in new ways. It should be said that for others, it made these problems indecipherable. The second is the adoption and adaptation of data science methods and approaches developed in one context into another. The third is the development of interactional expertise (Collins and Evans, 2002), or having enough expertise to talk meaningfully about a practical skill or expertise, but without being able to actually practice it (Collins 2004). In our fieldwork, this meant developing the capacity to “think like a data scientist” as a way of bridging knowledge boundaries and enculturating researchers. Fourth is the practice of institution building around data science, in which participants and organizations and those in the broader communities engage around issues that matter for the future of data-intensive science, such as reproducibility, open science, pedagogy, and data science studies or the social, ethical and organizational dimensions of data science.
Collins, H. E., Evans, R., & Gorman, M. (2010). Trading zones and interactional expertise. In Trading Zones and Interactional Expertise: Creating New Kinds of Collaboration (pp. 7-23). The MIT Press.

Collins, H. and Evans, R. (2002) The Third Wave of Science Studies: Studies of Expertise and Experience. Social Studies of Science 32 (2): 235 – 296.

Collins, H. (2004) Interactional Expertise as a third kind of knowledge. Phenomenology and the Cognitive Sciences 3: 125. doi:10.1023/B:PHEN.0000040824.89221.1a

Conway, D. (2013) The Data Science Venn Diagram. http://drewconway.com/zia/2013/3/26/the-data-science-venn-diagram

Galison, P. (1997) Image & logic: A material culture of microphysics. Chicago: The University of Chicago Press.

Moore/Sloan Data Science Environments, http://msdse.org/environments/

Authors

• Brittany Fiore-Gartland, eScience Institute

• Anissa Tanweer, University of Washington