Expanding the Scope of Digital Labour Research

Digital labour studies have traditionally focused on low-skilled workers, such as food delivery couriers and internet influencers, to emphasize their precarity under informational capitalism. This presentation aims to broaden digital labour research by extending the labour chain to include lower-status, cutting-edge workers such as AI data annotators; incorporating high-skilled technological labour; and unveiling the critical, yet unseen skills and organisational frameworks that underpin the digital industry.

This talk is grounded in two empirical case studies. First, this talk examines China’s AI data annotation work. Data tasks – accumulating, classifying, and annotating are vital for AI development. Based on three years of ethnographic research and 154 interviews, this study illustrates how annotation work bridges the gap between human values and machine recognition.

Two factors are pivotal in fostering such human-machine complementarity: firstly, the necessity for annotators to continually adapt their “pedagogical skills” either through “upskilling” or “downskilling” in response to evolving AI technologies; secondly, the role of diverse intermediary organisations in facilitating skill adjustments, disciplining annotators, and addressing emerging labour issues.

The discussion then turns to highly skilled tech workers, critiquing the oversight of this new generation of tech workers whose roles complicate the understanding of technological advancements, algorithm transparency, and the disparity between skill levels. Drawing from Tongyu Wu’s forthcoming book, Play to Submission: Gaming Capitalism in a Tech Firm (2024, Temple University Press), this segment explores the “field of games” in engineering processes at “Behemoth” (a pseudonym for a top American tech company). This 13-month ethnography suggests that gamification could define the future of work for both high-skilled and low-skilled workers amid increasing capitalist demands for technological innovation and iteration.