Abstract

Artificial intelligence is increasingly embedded in the everyday operations of public authorities, alongside growing institutional investment in AI factories across the European Union, envisioned as shared infrastructures operating at scale. While these initiatives emphasise preparedness, reuse, and coordination, they often rely on an implicit confidence in the stabilisation of data categories. This paper conceptualises dataset annotation as infrastructural and anticipatory work through which administrative categories are aligned for later deployment, even as their consequences remain temporally deferred across organisational contexts. Drawing on sociotechnical perspectives on classification and infrastructure, the paper develops an account of how such consolidation complicates the attribution of responsibility when classificatory effects surface over time. In doing so, it offers a temporally oriented understanding of data preparation as a site where governance is shaped through infrastructural arrangements extending beyond the moment of system design.

Download

Share

COinS