http://www2004.org/proceedings/docs/1p491.pdf

While searching for a relevant topic to do my last blog on, I came across this paper which couldn’t be any more applicable to class discussion and web blogging. Several researchers from IBM and MIT decided to develop an algorithm to model the dynamics of information diffusion through blogging. In their proposal, they discussed two types of blogging:

1) “chatter” or an ongoing discussion usually characterized by the author of the blog

2) “spikes” or a short term discussion usually stimulated by outside events found in the news, media, etc.

They also characterized four types of individuals who blog based on their posting behavior over a given amount of time. With these characteristics in mind, Gruhl and his group were able to construct a model, drawing upon similarities between the spread of information in a network and the spread of disease among a population. However, they found that a disease propogation model is not appropriate in information diffusion through blogspace because many topics can propagate without becoming “epidemic”.

The researchers go on to discuss other ways to analyze the situation, using ideas that included game theory approach, threshold models, power law networks - many topics discussed in class. Eventually they settle on using an independent cascade model for their study adding in a few extensions which account for behavior observed in online blogging. These include span of attention (the fact that people won’t read more than a few blogs at a time), stickiness (a particular topic arousing more interest than another), and multiple posts (authors publishing several posts on the same topic).

Being able to effectively model the nature of online blogging can help blogging sites (news sites in particular) filter out unrelated “spikes” in the posts so that crucial news blogs related to real-world events can be found more readily. These blog models can also enable companies to perform self-assessments that would otherwise be too costly to do.