The Dynamics of Competing Cascades in Social Media: Applications to Agenda Setting
With the rise of fake news, many researchers are studying how different information cascades on social media compete with one another and more importantly, the methodology of how to contain such information cascades of fake news. Here, they view the spread of information as an epidemic model. The strategy that they use to contain a spread of information is what they call “counter-contagion”. This strategy, in summary, “releases a stronger contagion into the network that will compete with and eradicate, the first contagion.” In this article, I will discuss the algorithm they develop for counter-contagion.
There are 2 basic steps of the algorithm that they develop:
1. Determine the strength of the counter-contagion
To do this, they calculate the rate of propagation of the information cascade that they want to stop using the SIRS model. Using that information, they come up with values
2. Identify which nodes to inject the “counter-contagion” with
The nodes that are chosen are nodes with a high “k-shell” value. Nodes with a high k-shell value no only “have many connections
but also are connected well-connected nodes.”
Once employed, the counter-contagion will overpower the “other virus,” and cause people to talk about the counter-contagion rather than the original contagion. They have not been able to entirely quell discussion of the weaker contagion, but it is significantly reduced.
An example of a counter-contagion in action (taken from this same article):
We learned about information cascades in class. This is the topic of information cascades applied to a very contemporary issue. They view the spread of information on social media in an epidemic model, another model which we learned in class. The strategies in the algorithms make sense, as they take advantage of nodes who are big “influencers” and are connected to lots of nodes who are also big “influencers”. We learned about strategies similar to this in class with things such as page rank, and again, information cascades.
http://people.cs.vt.edu/naren/papers/wsdm2014-difnet-competing-cascades-agenda-setting-camera-ready.pdf