Deliberative Mini-Publics, Artificial Intelligence and Scalability: Interrogating the Concept of Scale

It is increasingly common to encounter the claim that artificial intelligence can help scale processes of public deliberation. This paper seeks to problematise this claim by analysing the concept of scale. It does so by arguing that scale can be disaggregated in four ways, categorised as ‘scaling up’, ‘scaling out’, ‘scaling across’ and ‘scaling impact’. Respectively, these facets show that scale can relate to: (a) the level of governance a deliberative process occurs at, (b) the number of citizens involved in a deliberative process, (c) the number of deliberative processes occurring across different political issues and (d) the degree of impact the deliberative process has on the political system. An analysis of these facets of scale reveals that the pursuit of each faces unique practical barriers and normative trade-offs. It is argued that this analysis problematises the claim that AI can help scale deliberation in two ways. Firstly, it reveals that the claim lacks analytical rigour, and a deeper vocabulary is provided to support more grounded discussions on the kind of scale different AI applications may support. Secondly, it shows that the claim problematically and simplistically frames the issue of scale as one largely susceptible to a technological fix, obscuring the deeper social, economic and political barriers to scaling deliberation, and the complex normative terrain that arises when pursuing scale. This analysis can guide future empirical and theoretical explorations of the role of AI in scaling democratic innovation.