Community Structure and Information Cascade in Signed Networks
Source link: New Generation Computing Volume 32, Issue 3-4 , pp 257-269
http://link.springer.com/article/10.1007%2Fs00354-014-0404-7#
Apart from the rule we learned about information cascade in chapter 16, there may be relationship between cluster signed network and information cascade according to the newly study by Mahsa SHAFAEI and Mahdi JALILI in Sharif University of Technology.
According to their study, Social networks are different from many other real-world networks in that they show a community structure. The individuals in a social network are often organized in distinct communities with dense intra-community connections and sparse inter-community links. Community structure in a simple network has a distinct consequence on information cascade. It has been found that the more is the clustered structure of the network (which can be quantified by measures such as modularity index, the less the cascade depth. Also, diffusion from one cluster to the other depends on their density, such that the more homogeneous a network is, the more the cascade depth. In other words, it has been shown that the higher the density of clusters in a simple network, the more difficult the information spread in the network. Adopting various diffusion models, it was shown that for the same degree distribution, clustered networks have less cascade depth than those with less clustered structure. Signed social networks have community structure as well; however, often a different definition for communities in such networks is adopted. Further to have dense intra-community and sparse inter-community links, they often follow the social balance theory. In such communities, the ratio of intra-community links with negative sign and inter-community links with positive sign is minimal.
They investigate the relation between the community structure of sign networks and the cascade depth, and find a significant correlation between them. They study information cascade in signed social networks based on a game theoretic approach. They investigate the role of the number of initial adopters (i.e., the individuals that initially adopt the technology) as well as community structure of the network in the information cascade. Initially, a number of randomly chosen nodes are given the technology (or product). Then, based on certain rules, their neighbors decide whether to adapt the technology. This process is repeated until no further changes happen in the network and the nodes reach steady-state in their strategies. Finally, the number of nodes that has adopted the technology is extracted, which indicates the cascade depth; the more the number of such nodes, the more effective the cascade (i.e., the more the depth of the cascade).
And the finding shows that in model networks, increasing the inter-cluster links with positive sign decreases the cascade depth, while increasing intra-cluster negative links increases the cascade depth.
However, there’re some limitation in this study. One is the number of real networks considered in this work. Since they did not have access to enough datasets of real-world signed networks, they sampled from available signed networks. It is worth mentioning that the obtained correlations could be a mixed effect of having the same parent network used for sampling, although adopting the random walk strategy for sampling minimizes this effect. In order to confirm this findings with more confidence, the experiments should be performed on more real signed networks from diverse fields.