Reaction:Balancing Systematic and Flexible Exploration of Social Networks

I have seen quite a few visualizations that deal with social networks. The problem I have always faced with these visualizations is of information extraction. Due to the vastness of the social networks visualizing them becomes a big challenge. Even if you succeed in visualizing it, conveying the information to the readers or users is the toughest part. As there are so many nodes and links in the final output that it practically becomes impossible to differentiate between two elements.

This paper deals with exactly the same problem discussed above. I think it is an excellent article that shows how to effectively convey information while visualizing a social network. While visualizing a social network, the focus on relationships instead of just the individual elements; how the elements are put together is just as important as the elements themselves. So this paper talks about ‘Social Actions’ which is a social network analysis tool which balances systematic and flexible exploration. SocialAction applies attribute ranking and coordinated views to identify extreme-valued nodes. Social network analysts seek to uncover two types of patterns in networks: (1) those that reveal subsets of nodes that are organized into cohesive social groups, and (2) those that reveal subsets of nodes that occupy equivalent social positions, or roles.

The most interesting part of the paper comes when the author talks about how SocialAction actual helps in making the visuals more effective. When networks become large, ordered lists become quite long and network layouts become illegible. SocialAction alleviates this problem by allowing users to aggregate nodes based on link structure. This allows analysts to compress a network or examine communities that are of interest. Some social networks contain multiple types of links. SocialAction allows users to systematically iterate through them while maintaining node layout stability. A matrix overview is also provided to help discover patterns across different link types, such as temporal evolution.

Rest of the paper deals with different ranking styles and filtering based on the ranking patterns for example multiplex ranking, aggregate rankings for cohesive subgroups and comparison of rankings with scatterplots. Overall I feel it was a very informative paper. And definitely would like to keep few things in mind before visualizing social networks.