AI-Mediated Democratic Innovations in Contested Contexts: Participation, Constraint, and Resilience in China

Although democratic innovations are typically introduced to renew democraccy, they have also increasingly appear in contexts marked by political contestation, shrinking civic space, and digital disruption. This paper therefore examines how artificial intelligence (AI) mediates citizen participation in China, a country that has experimented with democratic innovations such as participatory budgeting and publoc hearings. In doing so, this paper seeks to delve into the intricaies and complexities of China’s digital participatory governance systems and considers what such cases reveal for broader debates on democratic innovations under pressure. Specifically, this paper asks: How does AI mediate and shape citizen participation in contested and constrained contexts, and what implications does this have for democratic resilience beyond liberal democratic settings? Drawing on a sociotechnical perspective, the research views AI not as a neutral facilitator but as a mediating infrastructure that actively structures participation by filtering, classifying, prioritising, and translating citizen inputs into governance outputs. These mediating practices are embedded in institutional and political logics that define the boundaries of legitimate voice. Empirically, the paper combines a systematic mapping study of academic research, policy documents, and participation platform descriptions to identify dominant AI functions (such as sentiment analysis, issue classification, and triage), participation formats, and governance rationalities shaping their deployment. By examining AI-mediated democratic innovations in an authoritarian context, this paper contributes a comparative lens to shed light on participation under constraint. Rather than treating China as an outlier, the analysis speaks to wider dynamics increasingly visible in hybrid regimes and liberal democracies alike, where algorithmic mediation, administrative rationalisation, and political control intersect. The paper thus advances a more contextual and globally informed understanding of democratic resilience in the digital age.