Determining Credibility of Information Cascades found in Twitter
https://www.emeraldinsight.com/doi/pdfplus/10.1108/IntR-05-2012-0095
This article describes a method that can determine how credible the information generated from an information cascade is on social media. Before we begin, just to review the main concept, an information cascade is essentially when people who are connected in a network influence each other’s opinions, and thus influence the “opinion” of the overall network.
In this article specifically, the authors try to determine if information cascade found on Twitter is credible or not. In the study, the authors define an information cascade to be all the Twitter messages related to a specific news event. The authors would then apply machine learning to first decide if the information cascade (the corresponding Twitter posts/messages) correspond to an actual event. Afterwards, the authors then applied machine learning to decide if the messages from the information cascade are legitimate or not.
The authors conducted this study using a specific case on an earthquake that occurred in Chile in 2010 and how this news propagated throughout social media. They found that within 30 minutes, the word “earthquake” was being propagated throughout Twitter. Furthermore, they found that two effective ways in determining whether or not the information being spread through a social network was credible or not – the frequency of the terms and clustering of certain words. For the former, it makes sense, as in an information cascade, an individual is more likely to adopt an opinion if the number of others who have adopted it are high. In Twitter, if certain words are trending, then the users are more likely to use these trending words. The same logic holds for the latter – if a group of certain words (which can be associated with whether or not information is credible or not using machine learning techniques), then users will be more likely to adopt them, regardless of credibility. Instead, the users will adopt them due to its popularity, meaning the user will forget whether information he/she has and take on what the majority is saying (as in an information cascade).