Information Diffusion in Blog Space Network
Recently, our course is mostly focused on the topic about Network Dynamics. From my own view, the contents about information diffusion and cascading behaviors are intriguing and are extremely applicable for us to learn about the large-scale network we’re in. In the lecture, professors introduced the concept of “information cascade”, that is, the information that one infer from others’ choices may be more powerful than his own private information, thereby forming the decision of one person even with no regard of his own private information. Furthermore, with the swift development of technologies and rapid popularization of Internet, people can have interaction and communication with strangers from all over the world, breaking the physical barrier.
In the research paper Information Diffusion Through Blogspace, collaborated by researchers from International Business Machines Corporation (IBM), D.Gruhl, R.Guha, David Liben-Nowell and A.Thomkins, four authors specifically and thoroughly study the propagation of discussion topics from person to person through the social network represented by the space of all weblogs. The authors characterize and model the whole blog posting network by looking closely at two major components: the characterization of discussion topics, and the type of individual that participate into the discussion.
Before the start of each branch’s characterization, authors first refer to the widely-adopted and classical disease-propagation modeling of information network. Under such modeling, the network follows the power law and the probability that the degree of a node is k is proportional to k^−α , for a constant α ranging between 2 and 3 (also covered detailedly in the textbook Chapter 18 about POWER LAWS AND RHICH-GET-RICHER PHENOMENA). However, such an ideal hypothesis contradicts with facts from real blog networks: “many topics propagate without becoming epidemics”. To construct a more realistic modeling of the network, authors further analyze two other fundamental models in information diffusions: Threshold Model and Cascade Model as we introduced in Chapter 19 CASCADING BEHAVIOR IN NETWORKS. Authors establish assumption that sharing discussion of a new and interesting topic with others in one’s immediate social circle may bring pleasure or even increased status.
In the studying of general discussing topic structure, researchers differentiate between the internally driven, sustained discussion— chatter, and externally induced sharp rises in postings — spikes. Consequently, the classification of topics are easy to conduct :
|
Just Spike |
Topics which at some point went from inactive to very active, then back to inactive. |
|
Spiky Chatter |
Topics which have a significant chatter level and which are very sensitive to external world events. |
|
Mostly Chatter |
Topics which were continuously discussed at relatively moderate levels through the entire period of our discussion window, with small variation from the mean. |
The second major component of the blog network the users. The authors directly lists four categories of individuals based on the users’ blog posts statistics : RampUp, RampDown, Mid-High, and Spike. The most active user with great portion of topics are classified as Spike, then with decreasing activity and participation, users are gradually falling into Mid-High, then RampDown, finally RampUp.
One major conclusion that authors reach after analyzing the transmission graph is “Fanout by Individual”. By referring the number of follow-on infection generated by each individual (the influence of one individual to his outside network) as fanout. It’s clearly that as time pass by, most users leave the topic with less energy than it arrived, transmitting to an expected less than one additional person. Such decreasing popularity of one blog topics just impede researchers from applying epidemic model to the information network. However, in contrast to the decreasing fanout, another observation is resonance phenomena. That is, a massive response in the community is triggered by a minute event in the real world. Although the resonance is surprising and hard to predict, the instant and large-spread of information diffusion is also an unique characteristics in online networks, due to the efficiency of information sharing.
I think this article is an extension of the information diffusion and cascade behavior we’ve learnt in the class. In the course, we’ve only covered the dynamics of social network in reality lives, however, this one explore more about the network in another form (online and abstract network). In this article, we can learn more interesting conclusions only exist in such complex network system.