Understanding Public Opinion Evolution in Extreme Rainstorm Disasters for Resilient Governance: A BERTopic-Based Event Logic Graph Approach
This study employs a BERTopic-based Event Logic Graph approach to analyze multi-platform social media data from the "23·7" Beijing rainstorm, revealing that while public attention focuses on disaster impacts and emergency response, critical information regarding warnings, preparedness, and recovery remains underrepresented, thereby highlighting the need for full-cycle risk communication strategies to enhance resilient governance.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a city faces a sudden, violent storm, the physical damage is only half the story. The other half unfolds in the digital space, where millions of people share their fears, ask for help, and demand answers. This digital chatter is not just noise; it is a complex, living map of how a society understands and reacts to danger. For emergency managers and government officials, reading this map is essential. If they can understand how people move from hearing a warning to seeing the damage, and finally to seeking rescue, they can communicate more effectively and save more lives. The challenge lies in the sheer volume and speed of this information. Social media posts are short, messy, and change by the minute. Traditional ways of analyzing them often treat these posts like static snapshots, counting how many people talk about "rain" or "rescue" at a single moment. But this misses the most important part: the story. It misses how one event leads to another, how a warning transforms into a crisis, and how a rescue effort sparks a wave of public emotion.
To solve this, researchers Qiaoyan Ding, Cui Li, and Qiyu Tian developed a new way to trace the life of a disaster online. They focused on the extreme rainfall that hit Beijing and the surrounding region in late July 2023, an event known as the "23·7" rainstorm. Instead of just counting words, they built a dynamic model that connects the dots between different types of posts. They gathered nearly 97,000 messages from four major Chinese social media platforms: Weibo, Douyin, Zhihu, and Xiaohongshu. These platforms represent different voices, from real-time help requests on Weibo to emotional videos on Douyin and technical discussions on Zhihu. The team used a sophisticated computer program to read these posts and identify specific events, such as "heavy rain falling," "trains getting stuck," or "rescue teams arriving." They then looked for the logical threads that tie these events together, asking questions like, "Did the warning come before the flooding?" or "Did the rescue effort follow the report of trapped people?" By linking these events in a specific order, they created a visual network that shows exactly how public attention flows from the start of a storm to its aftermath.
The result of this analysis is a clear picture of how the public mind works during a crisis. The researchers found that attention is heavily concentrated on two specific phases: the immediate impact of the disaster and the emergency response. When the rain fell, the public focused intensely on the damage—flooding streets, damaged buildings, and the famous incident where the K396 train was trapped with nearly a thousand people. Once the damage was clear, the conversation shifted rapidly to the rescue, with massive attention on teams working to save lives and deliver supplies. However, the study revealed a significant gap in the conversation. Information about preparation, such as early warnings and safety advice, and information about recovery, such as rebuilding homes and restoring services, received far less attention. The public narrative was "heavy in the middle" but "light at the ends." People were deeply engaged with the crisis as it happened, but they were less focused on the steps taken to prevent it or the long road to recovery.
This imbalance is not necessarily because people do not care about prevention or recovery; rather, it reflects how the information flows. The study suggests that the current system of communication often fails to keep the public engaged with the full cycle of disaster management. The "light at both ends" pattern means that warnings are often ignored until the damage is visible, and recovery efforts are overlooked once the immediate rescue is over. The researchers identified specific "hub" nodes in their network—key topics like "Beijing Rainstorm Warning" and "Beijing Hit by Heavy Rain"—that act as bridges connecting the physical disaster to the social reaction. These hubs are critical because they control the flow of information. If a warning is not clear or if the link between the warning and the actual danger is broken, the entire chain of public understanding can falter.
The study offers a practical path forward for emergency management. By mapping these logical connections, officials can see exactly where the public's attention is likely to go next. Instead of waiting for a crisis to unfold and then reacting, they can use these maps to guide the conversation proactively. For instance, when a warning is issued, authorities could immediately link it to specific, actionable advice on what to do, rather than just stating that rain is coming. Similarly, during the recovery phase, they can actively highlight stories of rebuilding and restoration to keep the public engaged and confident. The researchers propose that by treating public opinion not as a static list of topics but as a flowing story with cause and effect, governments can build more resilient communities. This approach does not just help officials manage a crisis; it helps them understand the human experience of the crisis, ensuring that the right information reaches the right people at the right time, from the first drop of rain to the final cleanup.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.