Analysis Politics Twitter Visualisation — Snurb, 10 September 2010

This post comes as something of a postscript to my four-part series about the key themes of discussion under the #ausvotes hashtag on Twitter during the recent Australian election campaign (17 July to 21 August 2010 – see posts #1, #2, #3, and #4). In addition to looking at the content of those tweets, I also wanted to examine the networks of conversation which took place during that time. This builds on our trusty Twapperkeeper #ausvotes archives from between 17 July and 24 August again.

Those networks are created by Twitter users including @replies, of course – e.g. ‘@snurb_dot_info’ to get my attention. I need to point out two major limitations of looking at @replies in this way, though: first, not all @reply conversations will necessarily continue to include the #ausvotes hashtag in further tweets – one way of describing this is to say that where #ausvotes is missing from follow-up tweets, the users @replying to one another have stepped away from the crowd and begun a more private conversation (though still in a public space, unless they move to direct messaging). What I’m analysing in the following, by contrast, are only public conversations where the #ausvotes hashtag was retained – i.e. where users were talking to (or at) one another, but did so still with the wider #ausvotes audience in mind; we might understand this as a deliberately publicly performed conversation.

The second limitation is that I’ve retained (old-style) retweets in the data. Retweets on Twitter have traditionally been in the form

[additional comment, optional] RT @[username] [original tweet, possibly shortened]

- so they do include an @username element that looks just like an @reply. Now, retweets may be used simply to distribute someone else’s tweet to one’s own followers, and in that sense don’t constitute much of a conversation with the original poster; however, many users also retweet a user’s original message as they reply to it, in order to preserve the conversational context, as in

I don’t think so. RT @someuser Abbott will lose the election. #ausvotes

- in this latter case, clearly they are part of an ongoing conversation. And even a simple retweet could be seen as a somewhat conversational act – at the very least, it indicates to the original poster that the retweeter is listening, and finds with their views interesting enough to forward them to others in the network.

(Twitter‘s more recent form of retweeting, where the original tweet is forwarded in full and without adding the ‘RT @username’, and all the relational information exists only in the tweet metadata, can not be detected very easily from the Twapperkeeper data, and is not included in the following analysis. That may be just as well, since those plain retweets are less versatile than RTs and thus less likely to be used in a particularly conversational manner.)

So, having said all that, here’s what the @reply data generate. First off, a quick map of the overall patterns of interlinkage in the network, which I’ll go on to explain in some more detail. The circles symbolise individual Twitter users; the lines between them indicate @replies (curved lines where @reply exchanges tend to be unidirectional, straight lines where there is persistent two-way @replying between two users):

(It’s a bit spooky that the most central nodes in the network seem to be arranged roughly like the Southern Cross, the constellation on the Australian flag. That’s purely by accident, of course…)

What I’ve done here is to use the open source network visualisation software Gephi to arrange the nodes of the network according to their strength of interlinkage – more frequently connected nodes attract one another, nodes which don’t link repel one another. So, nodes which cluster closely together are talking to one another most persistently, and the key nodes in the centre are the most central nodes in the overall network. (I’ve also excluded from the visualisation itself any users who received fewer than 100 @replies – their sent replies are counted in the figures below, though. Note: in Gephi, this is achieved by setting up an ‘Attributes > Range’ filter for the indegree measure; the ‘Topology > In Degree Range’ filter removes any stats contributed by filtered nodes altogether!)

The node sizes here are determined by the measure of indegree – that is, by how many @replies each user received between 17 July and 24 August. And node colours indicate a statistical measure called ‘betweenness centrality’ – in simple terms, the extent to which a user acts as a central connector – an information broker – for others in the @reply network (see the quick definition here, or Wikipedia’s more detailed explanation). The redder a node, the higher their betweenness centrality (and the lines between two users inherit their colour from the person sending the @reply). Here’s a ranking of users by those two measures.

juliagillard 2996   latikambourke 6018.534
annabelcrabb 2380   annabelcrabb 2830.6
abcnews 2012   correllio 2531.22
latikambourke 1783   peterjblack 1979.731
australianlabor 1639   mikestuchbery 1961.653
tonyabbottmhr 1411   ibleeter 1941.673
wendy4senate 1171   philbellamyinc 1847.012
catherinedeveny 1148   sunriseon7 1605.263
senatorbobbrown 1119   unsungsongs 1260.26
greens 901   jeremysear 1157.249
trubnad 892   bridgetoflynn 1091.668
oldspice 823   mfarnsworth 1055.677
antonygreenabc 745   drwarwick 1051.535
greensmps 723   newtonmark 997.9389
mpesce 721   australianlabor 992.515
unsungsongs 716   grogsgamut 983.9376
liberalaus 701   ben_hr 877.7656
sunriseon7 692   nickhodge 853.9994
chaslicc 681   michaelbyrnes 851.5939
onepotchef 662   renailemay 814.749
getup 656   firstdogonmoon 799.9014
abcmarkscott 642   jdub 752.4107
michaelbyrnes 618   fakepaulkeating 751.6562
turnbullmalcolm 584   miltonfriedmans 749.0204

Clearly, there are some significant differences between those lists – Twitter users like Prime Minister GIllard (juliagillard) or Opposition Leader Abbott (tonyabbottmhr) might get tweeted at a lot, but themselves tend to use the service more to make unidirectional announcements than to engage in two-way conversation, and therefore show up here as having lots of incoming @replies, but not as major information brokers. Remarkably, for example, Abbott did not tweet at all between 17 July and 18 August 2010, and sent no @replies to anyone during the election campaign, so his betweenness centrality is zero (Gillard, who sent 21 @replies, is at 570).

Political parties and some news organisations are in a similar situation – ABC News (abcnews), as well as the Labor (australianlabor), Greens (greens), and Liberal parties (liberalaus) are amongst the Twitter users who get tweeted at frequently, but don’t necessarily reply as much. Other notables here are the Twitter accounts of Wendy Francis (wendy4senate), a Family First candidate for the Senate from Queensland who generated some notoriety (and plenty of outraged replies) through a number of utterly bigoted comments posted by her Twitter account – and later attempted to deflect the blame to her campaign staff – and similarly controversial columnist Catherine Deveny (catherinedeveny).

The betweenness centrality stats are necessarily different from this list – here, indegree (received @replies) and outdegree (sent @replies) both count, as does the user’s overall location in the network. From that perspective, interestingly, the most important information brokers in the #ausvotes @reply network are two journalists: Radio 3AW reporter Latika Bourke (latikambourke) – by some margin – and the ABC’s political gossip columnist Annabel Crabb (annabelcrabb). Both of them have come to some prominence on Twitter as a result of their live tweeting from press conferences and other unfolding campaign events, and are located very centrally in the overall network, as the graph below shows.

Most of the other highly (betweenness-) central members of the network seem to serve different roles, however: for the most part, they are independent political commentators of some note, rather than professional journalists. Included here are Twitter users such as my QUT colleague Peter Black (peterjblack), who was also the public face of Electronic Frontiers Australia’s campaign against the government’s proposed Internet filter; former “This Is Not Art” festival director Marcus Westbury (unsungsongs); and regular ABC The Drum author Malcolm Farnsworth (mfarnsworth) – to name just a few – as well as a number of less well-known (user)names. I’ll need to analyse this further, but my immediate reading is that these people each act as a kind of repeater station for local regions of the wider network: unlike Bourke and Crabb, they’re not central nodes overall, but central to their own communities (comparing this network of #ausvotes @replies with the underlying follower network on Twitter might be interesting in this context).

Finally, then, and as promised, here’s the graph above once again, but with usernames added (zoom in for full details).

In the centre, you’ll again find the Southern Cross of Gillard, Crabb, ABC News, and Bourke, as the four most tweeted-at members of the #ausvotes @reply network; Australian Labor and Tony Abbott are also close by. Other users with high indegree, such as Wendy Francis, Catherine Deveny, or (to a slightly lesser extent) Greens Leader Bob Brown (senatorbobbrown) are more peripheral, indicating close attention only from a subset of the overall network (Brown, for example is closely connected with the Greens party (greens) and Greens MPs (greensmps) accounts, since they all retweeted one another’s press releases.

Of the key information brokers (in darker orange and red), only Bourke and Crabb are truly central to the graph; many of the others are slightly more distant from the centre, and there may act as hubs for their local networks – the straight lines which indicate bi-directional @reply exchanges also seem to support that reading; to the left of the Crabb node, for example, Peter Black (peterjblack) clearly sits at the centre of his own network of @reply exchanges.

A few interesting outliers are also in evidence, by the way – towards the bottom of the network (appropriately, given this), for example, we see the Twitter account for the Sunrise breakfast show on Channel 7 (sunriseon7), which acts as a gateway into a small cluster of other Channel 7 outlets (vote7news, 7newsqld, and y7news); to the right of the map are a few other journalistic outlets – such as Sky News Australia (skynewsaust), close to its political editor David Speers (david_speers); News Ltd. commentary outlet The Punch‘s editor David Penberthy (penbo); and SBS News (sbsnews), linked to its chief political correspondent, Karen Middleton (karenmmiddleton). These, then, are mainstream media staff and organisations who haven’t managed to position themselves either as central targets or information brokers in the overall #ausvotes @reply network.

About the Author

Dr Axel Bruns leads the QUT Social Media Research Group. He is an ARC Future Fellow and Professor in the Creative Industries Faculty at Queensland University of Technology in Brisbane, Australia. Bruns is the author of Blogs, Wikipedia, Second Life and Beyond: From Production to Produsage (2008) and Gatewatching: Collaborative Online News Production (2005), and a co-editor of Twitter and Society, A Companion to New Media Dynamics and Uses of Blogs (2006). He is a Chief Investigator in the ARC Centre of Excellence for Creative Industries and Innovation. His research Website is at, and he tweets as @snurb_dot_info.

Related Articles


(3) Readers' Comments