AI in Deliberative Democracy: Empirical Insights on the Legitimacy of AI-Supported Deliberation Platforms

This paper presents an empirical examination of deliberative platforms to understand how Artificial Intelligence (AI) facilitates or challenges democratic legitimacy. This review contributes with an analysis of deliberative platforms powered by AI, offering a nuanced understanding of the potential and risks associated with employing AI to augment public discourse and decision quality in the digital age. We have examined 27 platforms aiming to facilitate deliberations that involve lay citizens online, that have public institutions among their “clients”, and include discursive elements (Tsai et al 2024). A survey of the organisations running the platforms was also made. The research on the effect of AI on the throughput dimension of legitimacy yields several key insights. For example, several platforms are addressing throughput legitimacy with AI features designed to maintain mutual respect, build consensus, synthesis, and cluster information in order to support the forming of well-informed preferences. However, AI does not only bring new opportunities but also new risks for deliberation when it comes to user agency, transparency and auditability of the AI, privacy and anonymity. For example, if the AI is not sufficiently explainable and transparent, and engineers the deliberation process on behalf of users (e.g. by clustering their opinions or by summarising conversations in a certain way), the principle of agency could be threatened and trust of the users undermined. The efficiency of AI might also impact on the levels of creativity and dissent and put aside minority views.