Methodological Considerations for Centering Workers’ Epistemic Authority in AI Research

Abstract

The global AI industry is fueled by hidden and precarized labor, yet worker voices remain largely absent from the research that studies them. This paper introduces Workers' Inquiry as a Research Methodology or WIRM, a participatory methodology designed to shift AI research away from extractive practices and toward co-production with workers. Grounded in Marxist Workers’ Inquiry, WIRM centers workers as community researchers who shape the design, interpretation, and framing of research. We describe how WIRM was developed and implemented in the Data Workers' Inquiry (DWI) project across multiple countries, highlighting how the methodology adapts to diverse contexts of economic, physical, and political precarity. Rather than presenting empirical findings, this paper focuses on the methodology itself: its theoretical roots, structural components, and practical execution. We offer three case illustrations to demonstrate the method’s flexibility and limitations. We conclude by reflecting on key challenges and tensions, and propose WIRM as a replicable framework for worker-led AI research.