COMMColloquium

Co-developing an Epistemic Object: An Ethnographic Study of Deep Learning Algorithm Development in Radiology

Bomi Kim, Visiting Fellow, Cornell University

2pm in 102 Mann

Reception to follow in the Hub

How do experts from different knowledge communities collaborate to develop a technology that is as much defined by what it not yet is as by what it is? In this two-year ethnographic study of a deep learning algorithm development project, we interpret DL algorithms as “epistemic objects” for their ever incomplete, unfolding character and for how they elicit questions and actions from actors. We find that with its strong generative properties, the DL algorithm repeatedly presents surprises to experts, from which experts learn which direction to explore further as they continue to train the algorithm. Through such a chain of surprise-response, the DL algorithm ongoingly unfolds into a technology with ever-expanding functionality. This paper sheds light on the new interplay and learning between the algorithmic object and human actors. 

Bomi Kim is a fourth-year PhD candidate at the KIN Center for Digital Innovation, School of Business and Economics, Vrije Universiteit Amsterdam. She conducts phenomenon-driven, ethnographic research on how emerging digital technologies influence work and organizations using practice and process perspectives.