‘Chucking Their Stuff into ChatGPT and Hitting Send’: AI Slop in the Administrative State

Most research on AI in the administrative state focuses on its use by officials. This paper addresses a comparatively neglected issue: the impact of the use of AI tools by the public themselves. Drawing on a qualitative survey with over 500 UK civil servants, it argues that AI use by the public is increasingly disrupting the front-line of the administrative state. The argument is put in three parts. First, as the administrative state increasingly relies on text-based inputs, officials use correspondence as a proxy for individuals – a longstanding assumption of ‘autographic’ authorship that the diffusion of generative AI tools unsettles. Second, the problem can be characterised as 'AI slop': AI-generated correspondence with high volume and velocity that adopts a particular voice in interactions with officials. Third, this ‘AI slop’ causes three forms of disruption: to the signals officials draw from correspondence, to trust between officials and the public, and to systems ill-equipped to deal with it. The paper concludes by arguing that the status quo itself fails to adequately meet the needs of many people accessing services, and that systems need redesigning not only for a society with AI slop, but in ways that address the drivers of it too.