Why Neural Networks Are Not Bicycles: Some Problems with the “scot” Approach to Certain Technologies

The Social Construction of Technology (SCOT) (Bijker and Pinch 1987) approach is a widely known and influential framework within STS. Woolgar (1991) has argued that this (and related) approaches are ultimately dependant on “conventions of realist discourse” and that such an approach fails because it has not adequately addressed this dependency. In this paper I wish to show how the reliance on such realist discourse is foregrounded when we come to study certain technologies and conversely how such a reliance seems unproblematic when studying other technologies. I shall argue that not only does the failure to address such questions shows a compromise of the “radical potential” of SSK, it also conceals certain empirical questions that are central to the study of technology.