Affective counterpublics under constraint: emotion, platform governance, and AI ethics discourse on Chinese social media

This study examines how emotional expressions shape public discourse on AI ethics within China’s algorithmically curated and policy-regulated digital environment. Analyzing 20,037 Weibo posts from August 2023 to January 2025, we compare institutional and ordinary users’ engagement during key regulatory shifts and the launch of the domestic model DeepSeek-R1. Using LDA topic modeling and sentiment analysis, we find that institutional actors promote compliance-focused narratives, while ordinary users mobilize eight distinct emotions—particularly satire, outrage, and anxiety—identified through an extended NRC lexicon, to contest issues such as misinformation, plagiarism, and creative authorship. Notably, posts invoking existential or justice-oriented concerns receive up to 7.3 times more engagement than policy-aligned content, with Gold V accounts averaging 117.95 likes per post compared to 16.09 for Blue V accounts. The viral #AICheating campaign, which garnered 2.3 billion views, illustrates how affective discourse can pressure policy adjustments, contributing to national academic integrity reforms in 2024. We propose the concept of affective counterpublics under constraint—a theoretical framework that extends Fraser’s notion of subaltern counterpublics by integrating algorithmic resistance—to explain how emotionally driven grassroots mobilization operates within authoritarian platform governance. This reframes emotional expression as a form of embodied public reasoning capable of recalibrating policy attention under algorithmic suppression.