Critical Tools for Machine Learning: Working with Intersectional Critical Concepts in Machine Learning Systems Design

This paper investigates how intersectional critical theoretical concepts from social sciences and humanities research can be worked with in machine learning systems design. It does so by presenting a case study of a series of speculative design workshops, conducted in 2021. These workshops drew on intersectional feminist methodologies to construct interdisciplinary interventions in the design of machine learning systems, towards more inclusive, accountable, and contextualized systems design. The concepts of "situating/situated knowledges", "figuration", "diffraction", and "critical fabulation/speculation"were taken up as theoretical and methodological tools for concept-led design workshops. This paper presents the design framework of the workshops and highlights tensions and possibilities with regards to interdisciplinary machine learning systems design towards more inclusive, contextualized, and accountable systems. It discusses the role that critical theoretical concepts can play in a design process and shows how such concepts can work as methodological tools that nonetheless require an open-ended experimental space to function. It presents insights and discussion points regarding what it means to work with critical intersectional knowledge that is inextricably connected to its historical and socio-political roots, and how this reframes what it might mean to design fair and accountable systems.