She is currently working on a project calledRevising Ekphrasis, which uses advanced computational tools to explore connections between 4,500 English-language poems. You can find her online at LisaRhody.com and follow her on Twitter at @lmrhody.]
In last week’s post about social network analysis, I introduced NodeXL and its potential use for understanding online social networks. In this post, I want to focus on what we can learn from conference Twitter backchannel conversations, and how we can use software like NodeXL to improve the way we use social media to build computer-mediated scholarly networks.
Figure 1 (click to enlarge): A NodeXL network visualization of tweets from January 4, 2013 that include the hashtag #mla13.
The graph uses the Twitter images associated with each user id to represent an account (called an “actor” or “vertex” in the network) that sent a tweet, retweeted, replied to, or was mentioned in another tweet, and each “follows” relationship, tweet, retweet, reply, or mention creates a connection (represented by a green or blue line and called a “tie” or “edge”) between each Twitter user.
What can NodeXL show us about our conference Tweeting habits?
What we learn from the NodeXL graph is that the MLA Twitter network represents a tightly-bound community. Since the graph includes very few “isolates” (ie. tweets that come from people less closely connected to the group), the MLA network does not expand beyond those members already connected to the network. Marc mentioned this to me when he uploaded the graph. He explained that the MLA Twitter network represents an “in-group bounded community,” meaning that those who tweeted about MLA generally were read and retweeted by others with close ties to the MLA conference themselves. Few “outsiders” were visible in this network, in contrast to the many “isolates” found in more widely-discussed topics like brands or news events.
While it’s perfectly reasonable to assume that the Twitter backchannel during a scholarly convention may serve to form a more tightly-knit social network, the NodeXL graphs of the Twitter conversation at MLA 2013 also suggest that participation in the Twitter exchange did not broaden the community’s outreach. Sociologists distinguish between “bridging” versus “binding forms of social connections. The MLA Twitter network suggests it is used for bonding existing groups more than bridging to new connections. If the purpose of the backchannel conversation had been to strengthen existing ties, then the next step might be reaching out to connect to less-well-connected people, thereby extending the conversation to a larger community.
How can Twitter backchannel networks help us understand our community better?
Measures of betweenness centrality, for example, can show who creates the most connections in the network. In the case of the MLA graph from January 4th, the Twitter handles @MLAConvention, @rgfeal, @HASTAC, and @kfitz are central parts of the network because they act as hubs between disparate Twitter user groups. This is relatively unsurprising because the accounts are either the personal or professional Twitter presences of representatives from larger, pre-existing face-to-face networks, such as MLA and HASTAC. More interesting, though are the betweenness centralities of individuals, such as @briancroxall,@adelinekoh, and @mkirschenbaum. These are individual accounts not associated with the larger conference organizers whose tweeting was central to the interconnectedness of the MLA network as a whole.
Lists of words frequently used in the network of those who mention #mla13 are also detected in many of the MLA graphs. For example, when using the grouping algorithms to cluster Twitter users by their closest affiliations and most commonly used words, we can trace the communities within the network formed by the recurring use of particular vocabularies. In the case of MLA, session tags and workshop titles created smaller communities within the larger network.
By listing the top URLs used in Tweets from the entire #mla13network as well as from each of the smaller groups within the network, we can trace a pattern of interests, values, and sites of engagement that drew the most attention across the network.
What does social network analysis miss?
There are caveats to looking at network graphs. The graphs produced in NodeXL are not direct representations of whole conference networks, as they cannot capture real-time, face-to-face conversations that are not recorded online. Social media analysis is skewed toward those participants who are the most prolific, and more work needs to be done to understand the influence of single actors within the larger community. For example, the graph from January 4 includes the hashtag #wcw as one of the most frequently discussed terms for the day. In fact, what we know from experience of the conference is that there was a large amount of Tweeting going on in William Carlos Williams Society session, which made the #wcw hashtag appear so prominently. Even with more than 400 Twitter network participants, the sampling is still relatively small and so more easily influenced.
Still, social network analysis and visualizations produce much more than hairball graphs and can become a valuable activity to undertake — especially if you are interested in learning how to change, improve, and expand your own online scholarly communities. Future posts will help get you started on using NodeXL yourself: installing NodeXL, importing data from social media applications, and creating visualizations to share.
What networks would you like to better understand? Have you used social network analysis to understand your relationship to your peers online?