Mining public perceptions of shared autonomous vehicles in China : interpreting the temporal evolution of sentiment through urban heterogeneity

Abstract

The rapid deployment of shared autonomous vehicles (SAVs) in China raises pressing questions about public acceptance, yet systematic evidence on evolving public perceptions remains limited. Existing studies often focus on attitudinal surveys or single-platform analyses, overlooking the spatiotemporal heterogeneity of sentiment dynamics and their structural drivers. This study addresses these gaps by analyzing over 56,000 social media data from Sina Weibo, Xiaohongshu, and Douyin between 2022 and 2025. This study proposes an integrated framework combining various natural language processing (NLP) tools such as latent Dirichlet allocation (LDA), bidirectional encoder representations from transformers (BERT), and text network analysis (TNA) to capture thematic concerns and emotional trajectories. Six explanatory variables—policy support index, media campaign intensity, tech failure count, eco-transition input, AV test license count, and congestion index—are incorporated using Spearman’s correlation to diagnose city-specific sentiment drivers covering four pilot cities—Beijing, Shanghai, Shenzhen, and Wuhan—where SAVs have been commercially deployed. Findings reveal four salient themes: unemployment from automation, pricing and cost, safety and trust, and convenience and accessibility. Temporal sentiment streamgraphs highlight divergent urban trajectories: Beijing displays gradual but resilient optimism; Shanghai undergoes a biphasic reversal; Shenzhen demonstrates accelerating enthusiasm; and Wuhan experiences a sustained decline. Building on these patterns, this study advances a typology of urban SAV governance, distinguishing innovation-driven pioneering cities, policy-technology coupling cities, distrust-dominant cities, and governance-deficient cities. By linking discursive dynamics with structural variables, this study advances the methodological frontier of urban sentiment analysis and provides actionable insights for adaptive and context-sensitive SAV policy frameworks.