‘Affordances’ for Machine Learning

The field of machine learning (ML) has long struggled with a principles-to-practice gap, whereby careful codes and commitments dissipate on their way to practical application. The present work bridges this gap through an applied affordance framework. ‘Affordances’ are how the features of a technology shape, but do not determine, the functions and effects of that technology. Here, I demonstrate the value of an affordance framework as applied to ML, considering ML systems through the prism of design studies. Specifically, I apply the mechanisms and conditions framework of affordances, which models the way technologies request, demand, encourage, discourage, refuse, and allow technical and social outcomes. Illustrated through three case examples across work, policing, and housing justice, the mechanisms and conditions framework reveals the social nature of technical choices, clarifying how and for whom those choices manifest. This approach displaces vagaries and general claims with the particularities of systems in context, empowering critically minded practitioners while holding power—and the systems power relations produce—to account. More broadly, this work pairs the design studies tradition with the ML domain, setting a foundation for deliberate and considered (re)making of sociotechnical futures.