Information Cascade On Twitter
https://dl.acm.org/doi/pdf/10.1145/1935826.1935845
The paper “Everyone’s an Influencer: Quantifying Influence on Twitter” takes what we learned in class about information cascades to practice. It focuses on a specific social platform Twitter. It specifically tracks 74 million diffusion events on the Twitter follower graph over a two-month period in 2009 to look into the characteristics and relative influence of 1.6 million Twitter users.
Two of the most insightful facts related to the INFO 2040 class that I learned through this research is, “we find that the largest cascades tend to be generated by users who have been influential in the past and who have a large number of followers.” This links to what we learned in the lecture about the Cascade Model: how information from others shaped people’s beliefs and choices. The model we learned in class simplified the real-world situation in that people receive information in order and the first two people usually have more influence. This research turns this intuition more practically that people with a large number of followers are more influential and could generate their own information cascade. Despite these intuitive results, the finding that “predictions of which particular user or URL will generate large cascades are relatively unreliable” was really surprising to me. The reason for this fact is that word-of-mouth diffusion can only be successfully exploited by addressing a large number of prospective influencers and thereby collecting average impacts.
Moreover, the researcher discovered that, while the most influential users are also the most cost-effective in some instances, the most cost-effective performance may be attained by employing “ordinary influencers”—-individuals with average or even less-than-average impact. This is a really interesting fact that I again connected to what some models we learned during the class assumed “people with more connections in the network are more influential” while the Diffusion Model we learned recently assumed that “people care more about others’ behavior in their own networks and do not influence by people outside of their own networks”. After reading this research paper, I gained insight and evidence supporting both of the assumptions.