You are here

Web Science for Social Network Analysis

Athens.
After the rather unruly cultural panel, WebSci '09 has now moved on to the next keynote, by Noshir Contractor. His theme is the application of Web science to social networks, and he begins by noting some of the experimental mobile tools now available for social networking. The Web in general enables us to communicate and collaborate with any one at any time, but what is necessary are tools that enable us to identify who it is that we should be or want to be collaborating with. This is where social network analysis and Web science comes in.

This requires a better understanding of the theories about motivations for participation in multidimensional social networks; it needs us to develop the means of capturing, storing, and processing the data and metadata about social relations; it needs the qualitative and quantitative measures for understanding such data, and it requires the advanced computational infrastructure to analyse the sometimes extremely large datasets which emerge from all of this.

Noshir now shows some of John Kelly's snapshots of various national blogospheres - these exhibit a diverse variety of macro-level structures which require us to understand the underlying generative mechanisms which create these structures. Such mechanisms can variously be explained through theories of self-interest, social and resource exchange, mutual interest and collective action, contagion, balance (linking to complementary nodes), homophily (linking to similar nodes), proximity (technologies have not brought about the death of distance), and co-evolution, and Noshir's previous work explores some such theories.

Each of these motivations has its own generative mechanism, creating what Noshir calls structural signatures that point to the operation of these underlying mechanisms - what we need, then, are the statistical analysis techniques which detect structural motifs in the networks we observe. This moves from exploratory to confirmatory network analysis that can understand multilevel motivations for social network participation, and can help characterise the social drivers that create and sustain online communities. Such research must address large-scale networks in various fields (social, business, etc.), of course - and social networks can then broadly be characterised as interested in exploring, exploiting,mobilising, bonding, or swarming.

In Web 2.0, then, there are multiple types of nodes and multiple types of relationships, which can be identified through relational data and metadata. It's all about such relational metadata, therefore, and we must build and explore the technologies required to capture and analyse such metadata. And where other major research projects require massive start-up investment, the Web, Noshir says in riffing off the well-known Mastercard ad, is 'priceless': the data are available to us already.

One example for this is a project that examined the US response to hurricane Katrina by analysing the situation reports created by various emergency response authorities. What emerges from this is that in the early days of this event, the focus was largely on Florida, where the hurricane first hit, and that oil companies in Louisiana and the Gulf were already preparing for the impact there. FEMA was central in the early days as well, but loses centrality as the crisis continued, while the American Red Cross became much more central: a clear indication of FEMA dropping the ball as new orleans was hit, and of the Red Cross stepping in to provide emergency help.

Another example is the study of interaction in the online roleplaying game EverQuest: here, there are four types of in-game communication - partnership, instant messaging, player trade, and mail, and Noshir's group has plotted the networks of interaction for each of these communication forms. The result is that people are incredibly selective about whom they connect with, and that transitivity (friend of a friend connections) exists in all four forms. People talk to those of similar age, and to people who are relatively geographically close. There is some gender homophily: men want to play with men, but women also want to play with men...

Such understanding of network structures can be applied to enable better communication and collaboration in online communities, and one example for this is a project which has improved communication amongst researchers studying the carcinogenic effects of smoking. This was based in part on a bibliometric analysis of the articles published by these researchers, and can be used to identify likely fellow travellers and future collaborators.

Technorati : , , , , , ,
Del.icio.us : , , , , , ,