Deciphering the laws of social network-transcendent COVID-19 misinformation dynamics and implications for combating misinformation phenomena

Mingxi Cheng, Chenzhong Yin, Shahin Nazarian, Paul BogdanScientific 2021 · DOI

Abstract

The global rise of COVID-19 health risk has triggered the related misinformation infodemic. We present the first analysis of COVID-19 misinformation networks and determine few of its implications. Firstly, we analyze the spread trends of COVID-19 misinformation and discover that the COVID-19 misinformation statistics are well fitted by a log-normal distribution. Secondly, we form misinformation networks by taking individual misinformation as a node and similarity between misinformation nodes as links, and we decipher the laws of COVID-19 misinformation network evolution: (1) We discover that misinformation evolves to optimize the network information transfer over time with the sacrifice of robustness. (2) We demonstrate the co-existence of fit get richer and rich get richer phenomena in misinformation networks. (3) We show that a misinformation network evolution with node deletion mechanism captures well the public attention shift on social media. Lastly, we present a network science inspired deep learning framework to accurately predict which Twitter posts are likely to become central nodes (i.e., high centrality) in a misinformation network from only one sentence without the need to know the whole network topology. With the network analysis and the central node prediction, we propose that if we correctly suppress certain central nodes in the misinformation network, the information transfer of network would be severely impacted.

Introduction

The COVID-19 pandemic was accompanied by a massive "infodemic," where misinformation spread rapidly across social media platforms. Traditional approaches to this problem often focus on Natural Language Processing (NLP) for veracity detection or network science to track user-to-user transmission. However, few studies have integrated these methods to examine how the semantic content of misinformation itself evolves over time. This research addresses this gap by modeling misinformation as time-varying complex networks, aiming to decipher the mathematical laws governing their evolution and developing tools to predict influential content without requiring full network topology.

Methodology

The study utilized a dataset of 60,798 pieces of misinformation collected from Twitter between March and May 2020, categorized into types such as unreliable, clickbait, bias, political, and conspiracy. Misinformation networks were constructed where nodes represent individual tweets and edges represent semantic similarity, calculated using Sentence-BERT embeddings and cosine distance. Three network formulations were analyzed:

  1. Daily Networks: Independent networks built for each day to track temporal changes in centrality.
  2. Cumulative Growth Networks: To study preferential attachment and node fitness.
  3. Network with Node Deletion Model: A novel approach that mimics shifting public attention by removing nodes that fail to attract new connections for three consecutive days.

Statistical Laws of Misinformation Popularity

A key finding is that the popularity of misinformation follows a log-normal distribution rather than a power-law distribution. This suggests that misinformation dynamics are multiplicative and resemble ecological systems where topics compete for the finite resource of human attention. Statistical tests, including Kolmogorov-Smirnov (KS) tests, consistently favored the log-normal fit across all categories, with $p_{KS} > 0.1$ for the log-normal model while the power-law model was rejected.

Cumulative distribution function (CDF) of mean popularity for 'Unreliable' misinformation, showing a superior fit for the log-normal distribution ($p_{KS}=0.920$) over the power-law ($p_{KS}=0.000$).
Cumulative distribution function (CDF) for 'Conspiracy' misinformation, further validating the log-normal distribution model ($p_{KS}=0.300$) as a better fit than the power-law.

Network Evolution and Structural Fragility

The analysis reveals that misinformation networks naturally optimize for rapid information transfer, evidenced by increasing degree and closeness centrality over time. However, this optimization comes at the cost of structural robustness. The networks are fragile; removing a small number of central nodes can cause the entire misinformation structure to collapse. Furthermore, the node deletion model demonstrated that the network undergoes expansion-shrink cycles as public interest shifts between different topics, such as moving from political conspiracies to medical rumors.

Predicting Network Centrality via Deep Learning

To facilitate real-time moderation, the authors developed a Deep Neural Network (DNN) framework to predict a tweet's future network centrality using only its text. By feeding BERT embeddings into a 3-layer DNN, the model achieved high accuracy in predicting which tweets would become central nodes. For 1-day horizons, the model reached 94.19% accuracy for degree centrality and 94.25% for closeness centrality. This demonstrates that the semantic and syntactic features of the text itself are sufficient predictors of a tweet's future structural importance, bypassing the need for expensive topological data.

Conclusion

This research provides a comprehensive framework for understanding the dynamics of misinformation. By identifying the statistical laws of spread and the structural vulnerabilities of misinformation networks, it offers new strategies for intervention. Future work should focus on modeling the interaction between true and false information networks and developing more robust theoretical models that account for rapid shifts in collective attention.

Summary updated

Abstract

The global rise of COVID-19 health risk has triggered the related misinformation infodemic. We present the first analysis of COVID-19 misinformation networks and determine few of its implications. Firstly, we analyze the spread trends of COVID-19 misinformation and discover that the COVID-19 misinformation statistics are well fitted by a log-normal distribution. Secondly, we form misinformation networks by taking individual misinformation as a node and similarity between misinformation nodes as links, and we decipher the laws of COVID-19 misinformation network evolution: (1) We discover that misinformation evolves to optimize the network information transfer over time with the sacrifice of robustness. (2) We demonstrate the co-existence of fit get richer and rich get richer phenomena in misinformation networks. (3) We show that a misinformation network evolution with node deletion mechanism captures well the public attention shift on social media. Lastly, we present a network science inspired deep learning framework to accurately predict which Twitter posts are likely to become central nodes (i.e., high centrality) in a misinformation network from only one sentence without the need to know the whole network topology. With the network analysis and the central node prediction, we propose that if we correctly suppress certain central nodes in the misinformation network, the information transfer of network would be severely impacted.
BibTeX
@inproceedings{cheng2021deciphering,
  author = {Mingxi Cheng and Chenzhong Yin and Shahin Nazarian and Paul Bogdan},
  title = {{Deciphering the laws of social network-transcendent COVID-19 misinformation dynamics and implications for combating misinformation phenomena}},
  booktitle = {Scientific},
  year = {2021},
  doi = {10.1038/s41598-021-89202-7},
}
Figure