Characterizing COVID-19 Misinformation Communities Using a Novel Twitter Dataset database and codebook

Memon, S. A., & Carley, K. M. (2020). Characterizing covid-19 misinformation communities using a novel twitter dataset. arXiv preprint arXiv:2008.00791. https://arxiv.org/abs/2008.00791

University of California, Berkeley

Researchers worked to create a method and analysis in order to characterize two groups on Twitter: misinformed users or users actively posting misinformation about COVID-19 versus informed users or users actively posting true information and countering misinformation about COVID-19. What was found was that the misinformed community was more dense and organized, and could possibly be spreading disinformation. Their results also suggested that a large proportion of misinformed users are also anti-vaxxers, and that informed users tend to use more narratives. 

Characterizing COVID-19 Misinformation Communities Using a Novel Twitter Dataset

With the data the researchers extracted, they analyzed the communities, bots, linguistics, network, tone, and vaccination stance of the tweets and users. 

Characterizing COVID-19 Misinformation Communities Using a Novel Twitter Dataset data analysis

A Twitter API was used to collect data from the weeks following March 29, June 15, and June 24 of 2020. They used certain keywords and hashtags such as "COVID19", "coronavirus", "wuhan", and "vaccine." From this data, tweets were randomly selected for annotation. For annotation, 17 categories were determined for tweets to be classified into. Tweets were then randomly taken from these categories for data analysis in order to have a diverse dataset of tweets about a range of topics. Four thousand five hundred seventy-three tweets were annotated from 3629 users. 

Characterizing COVID-19 Misinformation Communities Using a Novel Twitter Dataset data collection and annotation

- the misinformed community is more dense than the informed community
- the reply network has much greater communication between the two groups than the retweet or mention network 
- the percentage of bots in the misinformed community (19%) is higher than the percentage of bots in the informed community (14%)
- bots may indicate a spread of disinformation
- informed users use more narratives as they use more pronouns, functional words, and family-related words
- there is no significant difference in tone between the two communities; both are highly negative
- misinformed users tend to be more informal than informed users, but informed users use more swear words
- many in the misinformed community are anti-vaxxers with 41% being so and 22% being anti-vaxxer bots

Learn Before

Related

Learn After