The hidden human labour powering AI: Introducing the Fairwork Action Research Project

Artificial intelligence is often seen as a mirror of human intelligence, an attempt to replicate the processes that occur within a human mind. However, a different perspective is presented in the book Feeding the Machine, co-authored by Prof. Mark Graham. AI is described as an "extraction machine." When users interact with AI products, they typically only see the surface and the outputs it generates. In reality, the extraction machine absorbs vital inputs—capital, power, natural resources, human labor, data, and collective intelligence—and transforms them into statistical predictions, which AI companies convert into profits. This process requires control over material infrastructure, workers, and knowledge.

The talk introduces the Fairwork project, an action research methodology designed to hold companies within the AI production network accountable. It examines how the Fairwork methodology has successfully functioned in the gig economy, having scored nearly 700 companies to date. Fairwork works with platforms to encourage pro-worker changes to policies and practices. Guided by the Fairwork Principles, companies improve conditions for workers and develop safer, fairer businesses. As a result of Fairwork's engagement, 64 companies have agreed to implement 300 pro-worker changes, covering all five Fairwork Principles. These changes include ensuring minimum or living wages, GDPR-compliant data management, sickness insurance, contracts aligned with local legislation, anti-discrimination policies, the election of workers’ representatives, and collaboration with local workers' associations.

The talk further explores how this methodology will be extended to AI supply chains to compel companies to act more responsibly.