How to Save Lives With Microblogs? Lessons From the Usage of Weibo for Requests for Medical Assistance During COVID-19

Document Title

How to Save Lives with Microblogs? Lessons From the Usage of Weibo for Requests for Medical Assistance During COVID-19

Document Information

• Subject Area: Seeking help and disaster response via social media during crises
• Keywords: Crisis informatics, social media, help-seeking behavior, disaster response, COVID-19, Weibo, user behavior, digital volunteers, public health, human-computer interaction

Research Background and Issues

• Issues/Challenges:

  1. Social media platforms like Weibo were widely used by patients and their families to seek medical assistance during crises like COVID-19, but their functionality in such scenarios remains unclear.
  2. Compared to traditional emergency services, help requests on social media often lack organization, and the impact of platform features on information dissemination effectiveness requires further study.
  3. Current research often focuses on the overall patterns of information dissemination on social media, with limited attention to how users utilize specific microfeatures (e.g., super topics, hashtags).

• Significance:

  1. Understanding the use and effectiveness of microfeatures can improve the design of social media platforms, enhancing the efficiency of help-seeking information during crises.
  2. In-depth analysis of user behavior during public health crises can provide actionable design recommendations for emergency response agencies.

• Research Motivation and Related Work:

  1. Previous studies have primarily focused on general behavioral patterns on social media, rarely addressing the diffusion of help-seeking information and the specific limitations of platform features.
  2. Earlier research faced issues such as insufficient data coverage and keyword search bias, leading to the omission of many help requests.

Solution

• Research Methods and Techniques:

  1. Data Collection: Collected over 100 million Weibo posts related to COVID-19 and used a machine learning model (BERT-based classifier) to identify 8,395 help requests.
  2. Data Classification: Conducted qualitative and quantitative analysis of the Weibo features used in the requests and their limitations.
  3. Time Series Analysis: Applied Controlled Interrupted Time Series (CITS) analysis to evaluate the impact of the introduction and exposure of Weibo's super topic feature on user behavior changes.

• Innovations:

  1. Identified nine major features used by Weibo users for help requests.
  2. Introduced deep learning-based natural language processing techniques to improve the accuracy and coverage of help request identification.
  3. Explored the impact of informal volunteers and user self-organization on the efficiency of social media platforms in providing assistance.

• Implementation Steps:

  1. Used machine learning to tag and classify help requests in the Weibo dataset.
  2. Analyzed the usage patterns, dissemination effects, and limitations of each feature (e.g., super topics, hashtags).
  3. Examined the influence of super topic exposure (e.g., appearing on Weibo's trending list) on users' choice of posting methods for help requests.

Research Outcomes

• Findings and Results:

  1. Feature Usage Patterns for Help Requests:
     • Super topics (43.5%) and regular original posts (51.3%) were the primary channels for users to post requests.
     • Users adopted super topics relatively late, with many relying on original posts despite their limited reach.
  2. Feature Dissemination Effects:
     • The repost rate for super topic posts (24.37%) was significantly higher than that for regular posts (18.88%), highlighting the advantages of centralized management features in information dissemination.
     • Posts containing the mention feature were more likely to be ignored (lower repost rates), suggesting that the mention feature might divert audience attention.
  3. Feature Limitations:
     • Insufficient search and tracking capabilities made it difficult to locate help requests in regular posts promptly.
     • Misuse of features (e.g., incorrect use of super topics) negatively impacted the efficiency of help-seeking.
     • Privacy protection issues were evident, with sensitive information (e.g., phone numbers) often being exposed.

• Limitations and Improvement Directions:

  • Although the introduction of super topics improved information centralization, their adoption was slow and not widespread.
  • More user-friendly interface designs (e.g., to prevent misuse) and higher levels of privacy protection are needed.
  • Platforms are advised to develop dynamic tracking features based on enhanced text analysis and duplicate detection technologies.

• Future Research Directions:

  1. Explore the broader role of informal volunteers in crisis management.
  2. Analyze help-seeking behaviors of other social groups (e.g., non-COVID-19 patients).
  3. Promote similar super topic features internationally and evaluate their applicability across different cultural and policy contexts.

Conclusion

• This study provides empirical evidence for improving social media features by analyzing help requests on Weibo during the COVID-19 crisis. The research highlights the importance of centralized management tools and the limitations of current microfeature designs, offering directional recommendations for optimizing social media use during crises.