Measuring Web-Based Networks

The next session at AoIR 2005 is on 'Emerging Research Methods for Analysing Civic Engagement'. Kenneth Farrall from the University of Pennsylvania makes a start. His paper is co-authored with Michael delli Carpini.

Kenneth Farrall: Web Graph Analysis

Kenneth begins by outlining perceptions of the Internet as alternatively a tool for revitalising democracy, or for furthering its decline. There are various approaches to analysing its role in this, from content analysis and user surveys to social network analysis, which appears to be a very useful technique for analysing user engagement in the Internet. But such analysis is difficult in an Internet context as social network associations online are highly fluid, and Kenneth suggests that content analysis may be useful to address such problems. 

Content analysis extracts data from text, and applies analytical constructs to the data to extract a set of inferences whose validity can be assessed. This analysis also take into account the specific context, of course. Such an analysis can also be applied to the analysis of Web links, then - this is still context analysis since ultimately it continues to analyse text, and can take into account various contexts of text as well (e.g. link texts, domain boundaries, etc.). Graph theory involves a number of concepts - vertices (nodes), edges (links), degrees (in and out links), geodesic (shortest distance between two nodes), metrics (centrality, density, cliques).

Social network analysis can now be applied to this data. Instead of analysing people as nodes in networks, then, here the connections between Websites are being analysed, but nonetheless many approaches and findings translate - for example, Stanley Milgram's idea of small world structure networks exhibiting higher local clustering than would be expected from a random distribution continues to apply. What is important here, though, is that what Web graph analysis finds is the topology of Web connections, rather than the flow of user movement across the network - users might take different paths from site to site than could be expected from a graph of the closest possible connections between sites.

Through Web graph analysis it is possible to identify the key actors in specific issue spaces. However, the validity of this kind of study remains a key question. Are the Web graphs developed in the process truly representative of the issue networks in question? The way this form of analysis works is through analysis from the starting point of a small number of seed Websites, so to what extent are findings dependent on the choice of seeds?

Kenneth now presents a study he has performed for the Electronic Privacy Information Center (EPIC), which found EPIC to be the most central site in this issue network. But to what extent did this finding depend directly on using EPIC as a seed site? His study found that while the seed had a certain level of importance, it did not affect the results very strongly - so overall perhaps the theory is valid.

Kirsten Foot: Links and Bridges

 Kirsten Foot from the University Washington is up next, with a paper co-authored with Steven Schneider from SUNY IT. This work is also related to their forthcoming book Web Campaiging. She notes that civic engagement on the Web requires a definition of engagement first. The Websites themselves do count as political actions inscribed on the Web, of course, but there are also very many Web users who don't create sites or content themselves, but nonetheless use the Web for civic engagement.

Kirsten also addresses the importance of online structures in this context. These includes features, links, and texts which provide opportunities for users to associate and act, and mediate the relations between site producers and visitors. In essence, site producers construct 'bridges' (not necessarily links) which connect site visitors and other actors: cognitive as well as transversal bridges. Such bridges influence where they go and how they interpret other sites and actors, this enabling online sociopolitical action.

Cogniitive bridges employ references to other actors through text or images; these invoke cognitive processes to facilitate connections between a site producer, site visitor, and other actor. Transversal bridges, on the other hand, are created through outlinks which enable users to direct their browsers to move from one site to another; they go beyond cognitive bridges by enabling the direct movement from site to site. An example would be an endorsement, which in itself is a cognitive bridge; adding an outlink to the endorsee's site then adds a transversal link which can be followed directly by the user. Both kinds of bridges can be utilised for affiliation, differentiation, or opposition, and simplify the engagement in political action.

What is required then is the systematic analysis of both kinds of bridging, that is, both link analysis and textual analysis as a more comprehensive form of content and network analysis of Websites. This may require both structured (type of relation) and unstructured (rhetorical frame) analysis of bridging. This is a 'mid-range' approach to making sense of bridging, filling a gap between micro-level analysis of text and links which focusses on particular links, and a macro-level, issue crawler link mapping analysis. It operationalises civic engagement opportunities as online structures that are available to Web users, acknowledging the directionality of bridging and highlighting the pathways through which civic engagement is potentiated (is that a word?) by users.

Alex Halavais: Blog News Agendas

Alex Halavais from the University at Buffalo is the third speaker, presenting a coauthored paper with Jia Lin and Sarah Whitehead. He begins with a brief overview of media agenda-setting theory. This is now complicated by the arrival of blogging and other forms of discourse which are beginning to have an impact on the traditional media. So, today, who influences what, when?

Alex suggests two approaches. The first characterises blogs as having an alternative agenda, and for this study Alex and the team looked at some 258 political blogs and newspapers over the period of a good week, using tools such as Minipar and CatPac. It found that there are some clear differences in the persons, topics, and political organisations being discussed in blogs and the mainstream media - and therefore it appears that blogs do present an alternative, but an alternative which presents topics that are more prone to be debated (especially controversial people or issues).

My battery ran out at this point, but another interesting finding from the study was in the temporal analysis of blog posts and mainstream news stories - looking for example at the coverage of hurricane Katrina, it was evident that peaks in blog posting activity appeared to precede mainstream news (Website) coverage by one or two days. One explanation for this was that bloggers watch TV, Alex suggested: that is, blog posts are inspired by stories in the mainstream news, and in turn then taken up later in mainstream news Websites.

But that's where my batteries did run out, and I couldn't blog Clifford Tatum's final presentation - a study of Website networks in the Montreal Chinese-Canadian community. A useful application of Web graph analysis.