Machine Learning, Infrastructures, and Their Sociomaterial Possibilities

Abstract

Abstract Machine learning hinges on various sociomaterial substrates, from computers where data is processed to infrastructures that support the networks of algorithmic experts. What happens when we place focus on these sociomaterial substrates? This chapter explores three distinct consequences. The first involves placing greater focus on organizational forms as contexts, enablers, and constraints for developments in machine learning. The second involves a focus on sociotechnical infrastructures, observing how the coevolution of practices, affordances, and built systems shape the trajectories of machine learning. The third involves being attentive to the way structural inequalities are reproduced and recombined by machine learning systems into novel categories of difference.