AI4Democracy: How AI Can Be Used to Inform Policymaking?
Colleen McKenzie, Deger Turan
Originally written for IE University’s Center for the Governance of Change.
Abstract
LLMs offer new capacities of particular relevance to soliciting public input when used to process large volumes of qualitative inputs and produce aggregate descriptions in natural language. In this paper, we discuss the use of an LLM-based collective decision-making tool, Talk to the City, to solicit, analyze, and organize public opinion, drawing on three current applications of the tool at varying scales: union decision making, coordination within DAOs, and nation- state consultations. We highlight the ways in which current-generation LLM tools can help leaders understand the needs of their constituents, review what measures are necessary to mitigate the flaws in these existing tools, and explore what future progress in foundation models would be most beneficial for the progress of tools like Talk to the City. We conclude that rapidly advancing AI capabilities offer substantial potential for informing and refining the process of governance, but demand strong and careful governance to mitigate their risks and take full advantage of their benefits. If applied carefully and with an understanding of the social context of its use, AI-driven technology for democratic decision-making has the potential to support collective agency in ways that systematically feed back into AI governance and AI safety institutions, creating a virtuous circle of improving AI’s impact on society.