Towards Actionable Data Science: Domain Experts as End-Users of Data Science Systems

Abstract

As a wider range of organizations explore using data science systems, data science research has given growing attention to the role of domain experts. Most of this research still views data science systems as centered on the development of statistical models or algorithms by technical data scientists, with domain experts limited to the role of informers. Our paper turns attention to how domain experts mediate whether data science models or algorithms lead to action through their situated data practices. Drawing on ethnographic fieldwork and a pilot machine learning project at a craft brewery, we identify situations where the brewers’ data practices led to unreliable, incomplete data, and unpack how such data limited the effectiveness of data science activities. Extending research in CSCW and STS on domain experts’ data practices to the data science context, we aim to inform the design of data science systems that are more actionable for their end-users.