Topic Shift in Information Cascades on Twitter
The possibility of an information cascade is at its prime when individuals make decisions sequentially, with limited access to the information that framed the decision of the link that they are being influenced by. Social media platforms create a solid foundation for information cascades with their ability to disseminate information to mass groups of people in record time. While this rapid spread of information is often seen as positive due to the accessibility of these platforms, it is important to also consider that social networks foster an environment that promotes the spread of misinformation (for example, think of fake news during the 2020 Election).
The social networking service Twitter differs from other popular platforms such as Instagram or Facebook by limiting the lengths of users’ posts, with its 280 character limit, forcing users to restrict the content of the information that they publish. The platform also actively supports sharing content outside of your immediate social network (ex. unlike other platforms, your feed naturally includes posts from people you do not know or follow, it is typical for content to get thousands if not millions of interactions, etc.) These conditions make users of Twitter even more susceptible to information cascades.
Liana Ermakova, Diana Nurbakova, and Irina Ovchinnikova examine information distortion that occurs in information cascades on Twitter, particularly during the COVID-19 pandemic. They point out that their study differs in typical misinformation research as they focus on how information mutates through being repeatedly shared on a social network (almost like a game of telephone) rather than a spread of an initially fraudulent message. During the pandemic, users were drawn to information from public (usually political) figures and personal stories “because complicated medical texts deter lay readers”. Sharing medical information in this fashion (paraphrasing or omitting important information to fit the character limit), generates cascades where the probability to distort initial information increases. The study answered two main questions: 1) “What are Public Figure tweets on healthcare topics that generate information cascades?” and 2) “How does a transformation of the initial tweet involve misinformation?” The study used 10M English tweets (141,866 original tweets, the remainder being retweets).
Users look to Public Figures to evaluate and to approve the available medical information. Due to the nature of a) the Public Figures not being experts and b) the design of Twitter as a platform, the study finds that “distortion and misinformation appear due to oversimplification and distortion of logical links.” When sharing medical information, Public Figures and/or their networks will shift the topic of the tweet to politics or business. Information cascades are rarely evoked by healthcare professionals. However, healthcare professionals were able to terminate an information cascade by “providing relevant information and ending discussion.” This is interesting because the definition of an information cascade assumes that an individual has limited access to insight about how the decision they are being influenced by was made, but social media naturally gives to access to basic information that could provide that insight. For example, it’s likely that you know the profession of someone who you follow on Twitter, particularly if they are a public figure, and have access to information regarding their qualifications or influences. Even if a user is initially encountering the piece of information after it has already been influenced by external sources, they are able to view who originally published the post.
This study gives us insight into how information cascades can not only perpetuate false information, but also distort an initially true message, and the factors that can influence the topic shift in such a cascade.
Resource: https://www.aclweb.org/anthology/2020.rdsm-1.3.pdf