Abstract

Social media bots (SMBs), automated accounts that produce content and interact with humans by mimicking human behaviors, are increasingly prominent on social media across the globe. Although benign SMBs can aid news reporting or health communication, the research community puts more attention on malicious SMBs as they are distracting or deceiving the public through spreading spam and disinformation, which threatens democracy and civic engagement. While previous studies employ various methods in detecting bots on English-based platforms like Twitter and Reddit, knowledge about bots in other platforms is limited. This study focuses on Weibo, the leading social media platform in China, as S MBs are increasingly prevalent and indistinguishable from non-SMB users on this platform but current bot detection techniques do not perform well to characterize their activities. Besides, previous research in this field relies heavily on text but largely overlooks data in other modalities (e.g. visual, network, traces) that also mark the presence of Weibo S MBs. Therefore, by exploring hundreds of millions of Weibo posts, we aim to develop a comprehensive multimodal framework to answer the following three research questions: What’s the typology of SMBs on Weibo? What are the features of SMBs on Weibo? How do they behave on Weibo?

Principal Investigators