How Sociotemporalities Inform Accountability: The Case of Targeted Advertising

In this talk, based on a paper coauthored for the special issue What is Sociological About AI? (Social Science & Computer Review), I make two arguments. First, I challenge the AI field's preoccupation with technical interpretability. Sociologists have reframed the black-box problem productively, showing that AI systems and their impacts become understandable through their material and social dimensions, including the minerals powering data centres, the labour sustaining operations, and the organisational relations and policies meant to prevent harm. Second, I extend this by drawing on the sociology of time, a resource largely overlooked in AI research and governance. Drawing on a retrospective empirical study of transparency efforts in targeted advertising, I show how sociotemporalities, the narratives and structures of time embedded through the sociotechnical fabric, help explain how AI systems evolve, and shape how accountability is imagined and enacted. I close by reflecting on how attention to sociotemporalities can inform analyses of emerging AI technologies and open new possibilities for meaningful interventions in AI governance.