Critical Tools for Machine Learning or CritML is a project that brings together critical intersectional feminist theory and machine learning systems design. The goal of the project is to provide ways to work with critical theoretical concepts that are rooted in intersectional feminist, anti-racist, post/de-colonial research and actualize them in machine learning (ML) systems design. The project is also explicitly grounded in new materialist feminist theoretical perspective and understands machine learning systems and their design as sociotechnical processes.
CritML consists (so far) of a series of workshops that explore concepts of “situated knowledges/situating”, “figurations/figuring”, “diffraction” and “critical fabulation/speculation” and work with them as guiding principles in ML systems design. These workshops open pathways to designing more inclusive, contextualized and accountable ML systems, and reframe their design as a transdisciplinary, contextualized process. The premise of the work is that while computer science has developed sophisticated technical tools to improve machine learning accuracy and expand application fields, critical theories, particularly feminist, anti-racist, post/de-colonial work, have developed tools to address systemic bias and intervene in hierarchies of power in society. The workshop series introduce critical intersectional methodologies as potential design intervention tools and suggest creative ways to translate such methodologies into tools for machine learning systems design.
A couple of publications have come out that introduce and document critML process:
Klumbytė, Goda; Draude, Claude; Taylor, Alex (2022). Critical Tools for Machine Learning: Working with Intersectional Critical Concepts in Machine Learning Systems Design. 2022 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’22), June 21–24, 2022, Seoul, Republic of Korea. ACM, New York, NY, USA, 1-14. https://doi.org/10.1145/3531146.3533207 . – This paper documents the process of the workshops and also includes a detailed description of the exercises that readers can use in their own design processes.
Klumbyte, Goda, Draude, Claude, & Taylor, Alex (2021) “Critical Tools for Machine Learning: Situating, Figuring, Diffracting, Fabulating Machine Learning Systems Design”. In CHItaly 2021: 14th Biannual Conference of the Italian SIGCHI Chapter. ACM, New York, NY, USA, 1-2. DOI: https://doi.org/10.1145/3464385.3467475 . – This offers a brief description of a workshop that took place during CHItaly’21 conference.
More materials will be uploaded as the project continues.