Cinelli, M., Quattrociocchi, W., Galeazzi, A., Valensise, C. M., Brugnoli, E., Schmidt, A. L., ... & Scala, A. (2020). The covid-19 social media infodemic. Scientific Reports, 10(1), 1-10. https://www.nature.com/articles/s41598-020-73510-5

University of California, Berkeley

Researchers looked at activity about the COVID-19 pandemic on five different social media platforms: Instagram, Twitter, Reddit, YouTube, and Gab from the timeframe of January 1st to February 14th of 2020. In particular, they looked at engagement, interest, evolution of information, and diffusion of misinformation. Some results found were that Twitter has the highest engagement for both reliable and unreliable posts, but also is the most neutral in amplifying such posts. 

The COVID‑19 social media infodemic

Keywords were used to determine posts used for the study on these platforms. Such keywords include "coronavirus", "pandemic", "coronavirus_transmission", "coronavirusnews", and more. API's were used, except for Instagram, to collect posts using the keywords to search for such posts. For Instagram, researchers manually looked for posts. 
They ended up with 1,342,103 posts and 7,465,721 comments from 3,734,815 users. 
These posts were then separated into "reliable" and "questionable" based on classification by MediaBias/FactsCheck (MCFC) which classifies sources as "Conspiracy-Pseudoscience", "Pro-Science", or "Questionable." Posts classified by the MCFC as "Conspiracy-Pseudoscience" and "Questionable" were placed into the "questionable" category, while those classified as "Pro-Science" were placed into the "reliable" category.

The COVID‑19 social media infodemic data collection

To look at the diffusion of information, they used an epidemiological model. The equation $$I=[\frac{R_0}{(1+d)^t}]^t$$ was used to define their exponential model in which I is incidence, t is the number of days, $$R_0$$ is the reproduction number and d is a damping factor. They also used the SIR model, described by three differential equations: 
$$∂_tS=−βS · I/N$$
$$∂_tI = βS · I/N − γ I$$
$$∂_tR = γ I$$
β is the rate in which a susceptible population can be infected when coming into contact with an infected individual.  γ is the rate in which someone who is infected can recover. S is the number of susceptible, I is the number of infected, and R is the number of recovered. 

To measure engagement, linear regression was used.

The COVID‑19 social media infodemic data analysis

- Topics related to the COVID-19 pandemic were similar across all five platforms
- Number of new posts about the pandemic was calculated, and there was an increase after January 20, when the WHO first reported on COVID-19
- largest increase of number of posts for Gab was Jan 21, Reddit was Jan 24, Twitter was Jan 30, YouTube was Jan 31, and Instagram was Feb 5
- the number of posts relying on questionable and reliable sources grow at a linear rate inside each social media platform
- there is a much smaller proportion of posts from questionable sources than reliable sources
- there is not a different pattern for posts that come from reliable vs questionable sources
- average engagement is largest in Twitter for both reliable and unreliable posts
- Twitter is most neutral in amplifying questionable vs reliable posts

Learn Before

Related