Abstract
This essay reflects on the shifting attention to the “social” and the “cultural” in data science communities. While recently the “social” and the “cultural” have been prioritized in data science discourse, social and cultural concerns that get raised in data science are almost always outwardly focused – applying to the communities that data scientists seek to support more so than more computationally-focused data science communities. I argue that data science communities have a responsibility to attend not only to the cultures that orient the work of domain communities, but also to the cultures that orient their own work. I describe how ethnographic frameworks such as thick description can be enlisted to encourage more reflexive data science work, and I conclude with recommendations for documenting the cultural provenance of data policy and infrastructure.