Abstract: [Purpose/Significance] This study systematically reviews the research related to rumor detection and governance on social media, so as to expand ideas for improving rumor detection technology and governance modes. [Design/Methodology] On the basis of conceptual analysis, the latest and classic literature on rumor detection and governance both domestically and internationally are retrieved. The research framework is organized from two dimensions: automated rumor detection and multi-subject participation governance mode on social media platforms. [Findings/Conclusion] The automated detection methods of rumors have evolved from machine learning to deep learning and then to graph neural networks; artificial intelligence technology has brought new challenges to the rumor automated detection; in practice, a rumor response mode with the government, platforms, and users as the main entities has basically taken shape. [Originality/Value] In the feature induction of rumor detection, this paper takes social context features as a special dimension in the social media environment, and pays attention to the differences in features and methods across different platform types. Focusing on the multi-agent governance model, this paper incorporates platforms and users into the governance agents, clarifies the research trajectory and development trends in this field, and proposes feasible research directions for the future.

Keywords: Rumor detection, Multi-agent governance, Social media, Social context