Communication of Uncertainty about Preliminary Evidence and the Spread of Its Inferred Misinformation during the COVID-19 Pandemic—A Weibo Case Study

Jiahui Lu, Meishan Zhang, Yan Zheng, Qiyu LiInternational Journal of Environmental Research and Public Health 2021 · DOI

TL;DR

This study investigates how uncertainty in preliminary scientific evidence, particularly during the COVID-19 pandemic, fuels misinformation on social media. By analyzing Weibo data, the researchers found that ambiguous communication of early evidence and the stage of its release significantly increase the spread and engagement with misinformation, suggesting a need for more cautious and clear communication strategies.

What Problem Does This Paper Address?

The rapid dissemination of preliminary scientific findings, especially during crises like the COVID-19 pandemic, has raised concerns about their potential to generate and spread misinformation. While much research focuses on external sources of misinformation (e.g., malicious actors, media), this paper addresses a critical gap: the role of internal sources, specifically how individuals' interpretations of uncertain preliminary evidence can lead to misinformation. The authors highlight that during novel health crises, people may rely on "naïve theories of science" and a desire to reduce uncertainty, leading them to infer unjustified claims from incomplete or ambiguous scientific information. This study aims to understand how the communication of uncertainty surrounding preliminary evidence influences the spread of such inferred misinformation on social media.

Methods

The researchers employed a mixed-methods approach combining computational analysis of social media data with statistical modeling.

  • Data Collection: They collected 3439 original Weibo posts and 10,380 reposts related to the misinformation that pets could transmit COVID-19. The data spanned from January 2020 to May 2020.
  • Sentiment Analysis (Machine Learning): A key methodological contribution was the use of a machine learning approach, specifically pre-trained Chinese BERT models, to classify users' attitudes towards both the misinformation statement ("Pets can transmit COVID-19 to people") and the evidence statement ("Pets can be infected with the virus"). This involved creating human-coded datasets for training and then applying these models to the larger dataset.
  • Statistical Analysis: Non-parametric statistical tests, specifically the Aligned Rank Transform (ART) technique, were used to analyze the data. This method is robust for frequency-type data and unbalanced sample sizes, allowing for the examination of interactions between different variables.

Datasets

The primary dataset used in this study consists of Weibo posts and reposts collected between January 2020 and May 2020, focusing on the misinformation regarding pets transmitting COVID-19. This dataset, comprising 3439 original posts and 10,380 reposts, was analyzed to understand user attitudes and engagement patterns. The paper does not introduce a new benchmark dataset but rather utilizes a real-world social media dataset for its analysis.

Key Insights and Ideas

  • Naïve Science Understanding Fuels Misinformation: The study posits that individuals, lacking deep scientific knowledge, tend to interpret preliminary evidence through their existing beliefs, leading to the generation of misinformation. This is particularly potent when evidence is uncertain.
  • Uncertainty as a Catalyst: Communicating uncertainty about preliminary scientific evidence doesn't just fail to prevent misinformation; it actively facilitates its spread. This suggests that ambiguity in scientific communication can be a breeding ground for false narratives.
  • Social Media Amplifies Uncertainty Effects: Platforms like Weibo, with their rapid sharing mechanisms, can exacerbate the impact of uncertain evidence, making it harder for accurate information to gain traction.
  • The "Seize and Freeze" Phenomenon: In times of crisis and uncertainty, people are motivated to quickly form judgments to reduce anxiety. Uncertain scientific evidence can trigger this by providing a seemingly simple, albeit false, narrative that individuals "seize" and "freeze" upon.

Key Results

  • Attitude Correlation: Users' attitudes towards the preliminary evidence and the misinformation were significantly correlated. Those who were ambiguous about the evidence were more likely to be ambiguous or endorse the misinformation.
  • Ambiguity and Engagement: Attitude ambiguity towards the evidence was associated with higher numbers of likes and reposts for misinformation posts. This effect was particularly pronounced when the evidence was in its early, uncertain stages.
  • Evidence Stage Matters: The "uncertain evidence" stage (e.g., when initial reports of pets testing "weakly positive" for COVID-19 emerged) saw a greater preference for misinformation and aversion to refutations compared to stages with no evidence or established consensus.
  • Interaction Effects: The interaction between attitude ambiguity and the evidence stage was significant, showing that ambiguous attitudes towards evidence led to more misinformation reposts, especially at later stages of evidence development, while suppressing refutations primarily at the uncertain evidence stage.

Limitations and Future Work

  • Limited Cognitive Insight: The study acknowledges that it could not directly measure users' cognitive processes, relying instead on inferring them from social media behavior. Future experimental studies are needed to explore these mechanisms.
  • Single Misinformation Topic: The analysis focused on one specific misinformation narrative (pets transmitting COVID-19). Generalizing these findings to other types of misinformation requires further research.
  • Contextual Constraints: The use of the ART technique limited the inclusion of covariates, meaning other factors (like emotional tone or user attributes) could not be controlled for.
  • Platform Specificity: Weibo's unique characteristics (asymmetrical relationships, open information flow) might influence results. Comparisons with platforms like Facebook or WeChat are suggested.
  • Future Directions: The authors suggest further research into other internal sources of misinformation, exploring the social-cognition mechanisms behind these findings, and developing strategies for building "misinformation-unfriendly" contexts through evidence framing and strategic social media use. They also emphasize the importance of health policymakers regulating the hasty communication of emerging evidence and using social media to communicate scientific consensus rather than uncertainty.

Summary updated

Abstract

The rapid spread of preliminary scientific evidence is raising concerns on its role in producing misinformation during the COVID-19 pandemic. This research investigated how the communication of uncertainty about preliminary evidence affects the spread of its inferred misinformation in a Weibo case study. In total, 3439 Weibo posts and 10,380 reposts regarding the misinformation of pets transmitting COVID-19 were analyzed. The results showed that attitude ambiguity toward the preliminary evidence and the stage when the evidence was first released with uncertainty were associated with higher numbers of likes and retweets of misinformation posts. Our study highlights the internal sources of misinformation and revisits the contextual perspective in misinformation studies.
BibTeX
@article{lu2021communication,
  author = {Jiahui Lu and Meishan Zhang and Yan Zheng and Qiyu Li},
  title = {{Communication of Uncertainty about Preliminary Evidence and the Spread of Its Inferred Misinformation during the COVID-19 Pandemic—A Weibo Case Study}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2021},
  doi = {10.3390/ijerph182211933},
}
Figure