Multo sa Makina: The Invisible Filipino Labor Force behind Artificial Intelligence

Abstract

Despite narratives of automation and innovation, artificial intelligence is built on human labor. The work of annotating, labeling, and rating the data that trains AI models is largely invisible to the public, yet it is foundational to every AI system in use today. The Philippines has long been recognized as a hub for outsourced labor, and news coverage points to its importance in the AI supply chain, yet its specific role as a source of AI data work remains largely understudied. This thesis examines the lived experiences of Filipino AI data workers through qualitative semi-structured interviews, drawing on heteromation, Fairwork's AI principles, and Weick's sensemaking framework to investigate their socio-economic positioning, working conditions, and how they construct meaning around their contribution to AI.

The study finds that Filipino AI data workers are positioned at the base of the global AI supply chain, shaped by a labor economy historically organized around service provision for other nations and constrained local employment alternatives. Dollar-paying remote work seems almost too good to be true. Workers treat it like a gold rush, enduring income instability, absence of social protections, and labor discipline, building products they do not know, to maximize the work while it lasts. In response to conditions marked by invisibility, uncertainty and precarity, workers engage in diskarte: an improvisational resourcefulness rooted in Filipino vernacular that functions simultaneously as agency, survival strategy, and sensemaking, while inadvertently normalizing the precarity it navigates. The experience of Filipino AI data workers is one of navigating a system built to hide them, finding ways to thrive within it, while the structure that demands that ingenuity, ultimately, remains unchanged.