Popularity Propagation Mechanism — A Case Study on Flickr
Link: http://www2009.eprints.org/73/1/p721.pdf
In chapter 18, we treated the popularity as a network phenomenon and had a deep insight into the distribution of pages with different in-links. However, the details of how popularity propagates in the social network aren’t clarified. In this paper, researchers take a further step and analyze the mechanism of popularity propagation in Flickr, an online social network for sharing photos. What is more interesting is that this work shows how the mechanism functions greatly depends on the network structure.
This paper aims to answer three basic questions: How far can popularity reach? How quickly does popularity spread? How does popularity flow through the network? In order to proceed the analysis, we need to have an overall vision of the network. The network structure of Flickr can be characterized as a small-world pattern with a strong local clustering with a small diameter. Based on this knowledge, this paper reaches some interesting findings. First they find most of the fans of a photo are located within limited hops, which demonstrates that information does not propagate widely in the Flickr social network. This finding might not be surprising if we take the network structure into consideration. Since the network is made up with small clusters, it would be more difficult for information to transfer across different groups, confining the incidence of the popularity. Further more, they also investigate how information propagates via social links. In other words, they want to clarify whether the social network would influence the spreading of the popularity or not. They find that social network plays a notable role in Flickr. There’s a clear increasing tendency to favorite-mark a picture if the number of friends who favorite-marked it before increases — a peer pressure happens here.
What are revealed in this paper are quite relational to what we discuss in class — information usually spreads inside a cluster and the more social links you have, the more likely you are to follow your friends’ decision. The paper also discuss the speed of popularity cascading, and three patterns of increment are summarized. However, there’re still something we can look into deeper. For example, is there also a “rich-get-richer” phenomenon in Flickr network? Or does the distribution of Flickr obey the Power Law?