Misinformation Spreading Like a Virus?
With COVID-19 at the forefront of most people’s minds, information cascades and diffusion patterns can help explain the topic shifts on social media relevant to the pandemic. In the article, “Covid or Not Covid? Topic Shift in Information Cascades on Twitter”, the authors discuss the presence of public figures, or individuals whose tweets are more likely to be shared with their followers. During the pandemic, many users on platforms, such as Twitter, turn to these public figures for relevant medical information and “follow personal stories of infected people who share unverified information”. Used in the context of spreading misinformation, the term “infected” is reminiscent of epidemics, with the “infected” information being the mutated medical information with omissions and paraphrase of jargon. Applying the Branching Process Model, each “infected” node (the individual with the misinformation) passes contagion to neighbors (i.e., their own followers and social networks) with a certain probability. As evident through epidemic models, each wave spreads to more nodes, as the normalization of mass communication has now eased the spread of misinformation on social media platforms.
Nonetheless, there still exist discrepancies between social networks and contact networks. The spread of information through a social network is determined by individuals’ choice in behavior, as evident through diffusion patterns and cascades on social media platforms. When a specific piece of information is passed to an individual on a platform, such as Twitter, the individual has two behaviors to adopt: accepting/passing on the information or rejecting the information. The decision whether to adopt a specific behavior or not can be dependent on other users’ choices. For example, if the individual sees that the majority of their network neighbors, which often times are close friends or followers, are interacting with this piece of information (such as retweeting it or commenting on relevant threads), then they have a greater incentive to follow such a trend. If individuals are convinced that the information is valid having seen the experiences of their network neighbors, they are more likely to adopt the behavior of their network neighbors.
In class, we learned that clusters were one hindrance to cascading; with respect to social media platforms, demographics could play a role in determining the formation of tightly knit communities. Since individuals tend to follow the behaviors of their network neighbors more than what the rest of the world is doing, homophily is a common phenomenon on social media. Individuals are more likely to relate to people of similar demographics/interest and tend to cluster together with these like-minded people. However, one downside of homophily is that it can act as a barrier to the spread of information between different clusters.
Homophily can help explain the patterns behind how COVID-19 information is spread on social media platforms. Political affiliation is one aspect that affects the circulation of information on social media platforms. Individuals of similar political affiliation are inclined to cluster together because of mutual interest, and conservative users tend to follow conservative news sources while liberal users tend to follow liberal news sources. The pursuit of biased information encourages the existence of echo chambers and promotes clusters with dense internal connectivity and weak external connectivity. This is evident in the image below, as the red and blue-colored clusters (signaling conservative and liberal nodes) are more densely intertwined within as opposed to between the clusters. This means that it is easier to spread information within each political group (either liberal or conservative) as opposed to between the groups (liberal to conservative, or vice versa).
This sentiment is further emphasized in the article, as the authors argue how “ordinary users exaggerate consequences of government decisions. They politicize and criminalize these actions shifting the topic to political and business disputes”. The politicization and distortion of scientific facts leads to greater polarization and further distances clusters from one another, preventing diffusion from occurring between different clusters of individuals. This further bolsters the point made in class that clusters block complete cascades from happening and that generally, when cascades come to a stop, a cluster is likely to be the explanation for it.
https://www.aclweb.org/anthology/2020.rdsm-1.3.pdf
https://www.cs.cornell.edu/home/kleinber/networks-book/networks-book-ch19.pdf
Picture: https://users.ics.aalto.fi/kiran/polarization-tutorial/