Politics of the prompt: Government in the age of generative AI

Abstract

This paper addresses the politics of the technique of prompting in machine learning, at a time when bureaucratic and democratic government is undergoing transformation. Drawing on the case of the UK government’s AI Redbox technology, we argue that prompting does not merely summarize documents, draft responses, or answer the user’s questions. Rather, the formulation of the prompt changes the politics of what can be said, what questions can be asked, what decisions can be made. Prompting represents a mutation in the exercise of political power, beyond the direct command and even the more indirect ‘action at a distance’ characteristic of liberal governmentality. We discuss three defining features of the governmental rationality of the prompt: describing a task; eliciting a desired response; and engineering a prompt. We examine how prompting constitutes a particular and novel form of the task in machine learning, and how political power becomes novelly expressed through the modelling of associations between inputs and outputs in the form of tasks. We explain how the prompt seeks to elicit desired behaviours from the model and why this elicitation changes the landscape of political action. Finally, we explore how prompt engineering is remaking expertise in algorithmic societies and conclude by addressing the wider implications of the prompt for the politics of language and critique.