The Politics of AI Systems Are Inextricable from Their Supply Chains: Public Values Versus the Digital Political Economy

Abstract

This article narrates the trajectory of a conflictual relationship between a public planning consortium and their AI technology vendor. Through an analysis of three years of email correspondence obtained via Freedom of Information Act request, I follow a team of transit planners from the U.S. state of Oregon who insisted that their vendor and its technology be transparent and accountable; conflicts emerged. I argue that this case demonstrates fundamental limits on the ethics and contextual appropriateness of AI tools that are dependent on data supply chains entangled with the digital political economy. These limits are enacted by such legal mechanisms as trade secrecy protections and non-disclosure agreements, as well as by the structural complexity and recursiveness of AI/ML model and data supply chains. Negotiations over the politics of the digital tool in question became articulated as conflicts over the provenance of the tool’s training data. By calling attention to how digital production relations are always situated within chains of dependencies, my analysis yields a more nuanced understanding about how the politics of AI-based tools are shaped in practice, and the terrain on which they might be contested and attempts made at re-configuration and alignment with public values.