A pre-print of the chapter is available here. And we promise to update you a little more frequently on our blog research in the future, too…

In keeping with the conference theme, “Media Leadership in Crisis, Disaster and Emergency”, we presented our research into the role of Twitter in the Queensland floods, but focussed here especially also on the intersections between mainstream and social media and the specific contributions of the Australian public broadcaster, the ABC, in this context. This also provided an opportunity again to show some of our research into the overall shape of the Australian Twittersphere (which we shared at the Digital Humanities Australasia conference in Canberra recently), and to examine the specific position of the ABC within it. More on that soon!

My contribution to the workshop was to present in some detail the research into uses of Twitter in disasters which we’ve been engaged in, and I’m including the two parts of my presentation below. There was a fair amount of interaction and discussion with the workshop participants during these presentations, but I’ve edited out those discussions from the audio attached to the slides, both because participants were hard to hear on the recording and because they may not want to have their questions and ideas broadcast beyond the workshop itself. Hope the recording makes sense without their contributions.

In particular, we examine the evolution of the Twitter user community and its activities over time – both in the short term over the fortnight following the February quake, and in the longer term through a comparison of the four major tremors between September 2010 and December 2011. These repeated events provide us with a useful indication of how Twitter use changes and evolves from one crisis to the next, building on the experience of its users in earlier events – but we do hope that there will be no further major earthquakes in Christchurch which add to this dataset…

In addition to the presentation itself, a full peer-reviewed paper will also be published as part of the ANZDMC proceedings – we’ll add it to this post and list it on our publications page once it’s available.

First off the rank: with one of our colleagues at the National Cheng Chi University in Taipei, Eugene Liang (whom Jean and I visited in October last year), I’ve just published a new article in First Monday which reviews some options for tracking user activities on Twitter. Our focus here is largely on doing so in the context of crisis communication, but much of what we have to say also translates to other applications.

My contribution to the paper is to outline what we’ve been able to do by using yourTwapperkeeper (and more details on this are all over this Website, of course), while Eugene takes us beyond that approach to discuss the work which he and his colleagues have done in developing a more comprehensive, custom-made platform for tracking Twitter activities at scale. Abstract below:

Tools and Methods for Capturing Twitter Data during Natural Disasters

Axel Bruns and Eugene Liang Yuxian

During the course of several natural disasters in recent years, Twitter has been found to play an important role as an additional medium for many–to–many crisis communication. Emergency services are successfully using Twitter to inform the public about current developments, and are increasingly also attempting to source first–hand situational information from Twitter feeds (such as relevant hashtags). The further study of the uses of Twitter during natural disasters relies on the development of flexible and reliable research infrastructure for tracking and analysing Twitter feeds at scale and in close to real time, however. This article outlines two approaches to the development of such infrastructure: one which builds on the readily available open source platform yourTwapperkeeper to provide a low–cost, simple, and basic solution; and, one which establishes a more powerful and flexible framework by drawing on highly scaleable, state–of–the–art technology.

A number of very interesting patterns are emerging from this. In addition to confirming some of the clustering tendencies which we already observed in the previous map (then based on data for some 550,000 users), I’ve also been able to overlay activity data for some of the #hashtags we’ve tracked over the past couple of years – so for the first time, we’re seeing those hashtags in the wider context of Twitter user networks in Australia.

Here’s my presentation, with audio from the talk:

And that’s in spite of Bligh’s (and Labor’s) very active engagement on Twitter throughout the campaign, of course, which I’ve documented over the past four weeks. I can’t help but see this as a sign that the Twitterati are already talking about Premier Newman. More in a final update on the weekend.

Feature image by ABC Open Wide Bay.

Before we get started, I’m also happy to report that our research to date has been featured in a great piece on the ABC’s 7.30 Queensland programme, which is available here, and embedded below:

Also, another brief reminder about the nature of the data we’re looking at for this study: we’re working off a dataset which contains tweets from and to all the candidate accounts we’ve been able to identify so far (drawing on the Courier-Mail’s Twitter list of accounts). Included in our dataset are all tweets which were either sent by those accounts, or @mention them (in the form of @replies or manual retweets).

I’ll get to the usual stats for leading politicians’ accounts in a minute – but given the significant shifts in public sentiment towards the LNP and its lead candidate Campbell Newman, let’s also have a look at some of the key moments in public discussion this week. To begin with, here’s the total volume of tweets to and from the candidates over the past seven days (I write this on Sunday, so our data for 18 March only includes tweets up to noon):

There are some obvious spikes during the week, which become even clearer if we look at the hour-by-hour activity levels:

We have some notable, isolated spikes, especially on 16 and 17 March. The major spike on 16 March, between 7 and 10 p.m., is driven by a public debate between the two candidates for the Premiership – and such Twitter spikes are a common phenomenon around candidate debates; we’ve also seen them during the last Australian federal election, for example.

The earlier spike, between 1 and 2 p.m., is more interesting: it’s provoked by a press conference by Premier Anna Bligh, who used it to present new allegations of misconduct against her opponent for the Premiership, Campbell Newman. And there’s a second, related spike the next day – on 17 March, Twitter activity between politicians and their followers spikes especially around 4-6 p.m., around a breaking news story: the announcement that the Queensland Crime and Misconduct Commission had not found any evidence of official misconduct by Newman in his previous role as Lord Mayor of Brisbane, and would not investigate further (but would further investigate donations to Newman’s lord mayoral re-election fund, made by property developers). Of the over 330 tweets we captured during those two hours, more than half contained the abbreviation ‘CMC’ in fact. It seems safe to infer that this story would have had a great deal of resonance also well beyond our dataset, and may have contributed significantly to the decline in support for the ALP and Premier Bligh (who had made some very public allegations against Newman).

These developments later in the week seem to have overshadowed the major story which dominated the start of the week – the ongoing debate around gay marriage and civil unions in Queensland. The substantial level of activity during Monday was driven largely by these themes: by a controversial ad against gay marriage from Katter’s Australia Party, by a GetUp! response ad featuring Katter’s gay half-brother, and by Campbell Newman walking our of his own press conference in order to avoid answering further questions about his stance on civil union legislation in Queensland.

The third major theme of the week – the release of the Queensland Flood Commission’s final report, around midday on 16 March – hardly even registers against these other themes, interestingly. We’ve yet to analyse the comprehensive dataset on the #qldvotes hashtag, which we’ve also been gathering throughout the campaign period – but if the patterns in the @reply conversations around politicians which we’re examining here are any guide at all, it looks like the Labor Party’s allegations against Newman, unsustained as they’ve turned out to be, have backfired quite substantially.

Keeping all this in mind, let’s have another look at how the leading political Twitter accounts’ stats are stacking up this week (and you might want to check out the corresponding graphs for the previous three weeks as well). There are a number of obvious shifts here: for the first time in the campaign, @TheQldPremier isn’t the most active Twitter account in this leadership group – the generic party account @QLDLabor has taken over, with nearly 170 tweets this past week. Deputy Premier @andrewfrasermp has slipped back a little, while Ashgrove candidate (and major hurdle to a Campbell Newman Premiership) @katejonesmp has maintained her level of activity. (Running out of competition, incidentally, it’s also surprising that former ALP Premier Peter Beattie’s @SmartState1 account has really dropped back – after nearly 100 tweets in week 2, and almost 60 in week 3, he’s now only managed just over 20…)

Overall, though, Labor accounts are still well ahead of their LNP counterparts. After a poor showing in week 3, the generic @LNPQLD account has ramped up its activities to a level approaching its week 2 volume of tweets again, and @Campbell_Newman has similarly nearly tripled his number of tweets this week (while following through on his stated intentions to terminate his brief flirtation with a separate, ‘personal’ account, @CD_Track). Even the deputy parliamentary opposition leader @TimNichollsMP has become more active again, if largely through retweets. (Parliamentary opposition leader Jeff Seeney remains deliberately absent from Twitter, as he told the ABC’s 7.30 Queensland programme this week: “I’m not that technologically advanced”.) But even so, the combined number of tweets from these accounts doesn’t even match the number of tweets by @QLDLabor alone – in terms of volume alone, ALP accounts maintain their supremacy on Twitter.

But such high levels of activity mean very little if there is no corresponding public response. Finally, then (I’ll skip the @reply network graph this week as this post is long enough already), here’s our usual leaderboard of tweets and @mentions by accounts captured in our dataset (which, to repeat, contains tweets to and from the 80-odd political accounts we’re tracking). As always, account names are colour-coded according to party (and while @JuliaGillard is obviously a Labor politician, she’s not a candidate in the Queensland election, so no colour for her here…).

The activity table, on the left, clearly shows the drop in tweets from @TheQldPremier, which was consistently near the top of that list in previous weeks. By contrast, Katter’s Australia Party candidate @RonWadforth is a new entry here, and other KAP politicians are also very visible here – perhaps this is related to the controversy around the party’s anti-gay marriage ads. The @LNPQLD account has also returned to the top 20.

It’s the other side of the ledger which is a better indicator of the overall visibility of these accounts, though, as it measures the number of @mentions (through @replies and manual retweets) they’ve received throughout the week. Here, it’s notable that after running neck-and-neck for the majority of the campaign, @Campbell_Newman has now pulled ahead of @TheQldPremier more decisively, receiving 500 more mentions this week (he was just over 50 mentions ahead over the same period last week). Both leaders also received some 50% more @mentions in total over the past seven days than they did last week – and on average, this is also true for most of the other accounts on the list; clearly the campaign is heating up. This top 20 is clearly dominated by the major parties; Katter’s Australia Party candidate @aidanmclinton is the only minor party representative here (again, no doubt related to the advertising controversy early in the week).

And if accounts are not related to the parties themselves, they belong to journalists and news organisations. If we keep in mind how we’ve gathered this dataset, this also tells us something interesting about how Twitter is now used by journalists during the campaign: remember that tweets will only be included in our data if they’re from candidate accounts, or @mention them – so tweets which @mention journalists are either from politicians back to journalists, or @mention both journalists and politicians in the same tweet (they could be manual retweets of journalists’ tweets which mention politicians, for example). This, then, demonstrates how central to the Twitter conversation about the election these journalistic accounts have become: they’re using the medium both to converse publicly with politicians, and to share news and observations about them – and their messages are widely disseminated by members of the general public who follow the campaign.

And that’s it for this week. I’ll post a final update next weekend, probably on Sunday once we know the final outcome of the election; good luck to all candidates participating in the poll, regardless of political persuasion!

Before we get to the week’s tweets, though, I thought it would also be useful to take stock of the campaign so far – from the start of the ‘phoney’ election campaign in late January through to the official start of the election period on 19 Feb. through to yesterday. Here’s the overall volume of tweets, and the number of unique users participating, on a day-to-day basis. Clearly, there’s a gradual growth in activities, boosted substantially on 19 Feb. as the election is officially announced. Over the past three weeks, we’re also seeing some obvious weekly patterns; activity on the weekends is markedly lower than on weekdays (and fewer users are participating, too). Additionally, the weeks themselves seem somewhat front-loaded – there appears to be more activity during the early weekdays than towards the latter part of the week. Perhaps this is an indication of the week’s pre-planned talking points being rolled out by each party’s political operatives?

In the background of the graph, I’ve also included the percentage of tweets on each day which were contributed by three different groups of users: the one per cent of most active contributors during the full seven weeks so far; the next nine per cent of still highly active users; and the remaining 90% of least active contributors. Very obviously, the top ten per cent – and amongst them, the top one per cent especially – dominate proceedings on each day; at the same time, however, the dominance of the lead users is also gradually and very slowly being pushed back, as a greater number and range of Twitter users begin to take at least some interest in discussing the election and start to post @replies to or retweets of the politicians. This appears to be a common pattern – well beyond Twitter itself – for public attention to election campaigns: it’s in the last few weeks of a campaign that the wider electorate, beyond the political junkies, actually begins to pay any attention.

But back to week 3 – and back to the principal political actors. Compared to our observations from week 2, which largely showed business as usual, week 3 of the campaign brings up some genuine surprises in activity patterns: at least for this week, the LNP opposition seems to have largely ceded Twitter domination to the ALP government. While @TheQldPremier Anna Bligh maintains her high rate of tweeting (sharing the load with her staff, it should be noted), and matches her activity figures from week 2 almost exactly, the overall @QLDLabor account has now also woken from its slumber, and was more than three times as active as it was in week 2 (we didn’t track it in week 1). Deputy Premier @AndrewFraserMP, on the other hand, dropped back, as did @katejonesmp, who contests the crucial seat of Ashgrove against LNP leader Campbell Newman; both only managed roughly half of last week’s Twitter work rate. Also continuing at a reduced rate (from nearly 100 to just under 60 tweets per week) is former Labor Premier turned media commentator Peter Beattie. Still, a good showing overall from the ALP, whose two major accounts now lead this group by some margin.

On the conservative site, the story is remarkably different. Last week, the @LNPQLD came second in this group, with nearly 120 tweets; this week (again, at least through to midday on Sunday), it managed only about a third of that number. Given the already very limited performance of the LNP’s candidate accounts, this is highly surprising: as I noted for week 2, it seemed as if the party had centralised its Twitter strategy, and used the generic party account as well as, to a lesser extent, that of its candidate for Premier, @Campbell_Newman, as the chief drivers of its Twitter campaign. If that was the case, something must have changed quite seriously in week 3. The malaise is also visible for that other account, in fact: @Campbell_Newman, posting a modest 34 tweets in week 2, only managed a paltry nine tweets in week 3 – an astonishingly low number for a political leader running for a major office (even if Newman himself had no interest in Twitter, you’d expect his campaign staff to keep the account ticking over at a higher volume, at least).

To further confuse things, you’ll also note the appearance of a new account in our little leaders’ group: the cryptically named @CD_Track. This has nothing to do with music playlists – the ‘CD’, it seems, stands for Campbell Newman’s chosen nickname and campaign slogan, ‘Can Do’ –, but represents an (apparently short-lived) attempt by Newman to present a more personal, personable image on Twitter. This could be seen as Newman’s very own ‘Real Julia’ moment, perhaps.

@CD_Track began on 5 March with observations of life during the campaign –

On the bus to toowoomba – free lollies and lemon slice were handed out !

–, but ceased to be updated only three days later, at least according to a tweet it sent to a reporter from the Gold Coast Bulletin:

@sueGCB sue – thanks for your help in getting going. I am going to use Campbell_Newman from now on and cease use of CD Track

Although its final tweet to @sueGCB denies any pressure from the LNP campaign team to shut down this second account –

– it is nonetheless not difficult to imagine campaign strategists’ unhappiness with the dilution of Newman’s online presence which this duplication of Twitter accounts presents. @CD_Track managed ten tweets during its brief existence (though none of them were particularly likely to have much impact on its followers’ voting intentions); even combined with @Campbell_Newman’s activities during week 3, those 19 tweets in total represent a significant drop from last week’s 34 @Campbell_Newman tweets, and they fail to come anywhere close to matching the Labor accounts’ much more substantial work rate.

The fact that the final major LNP account, that of deputy opposition leader @TimNichollsMP, returned to a somewhat greater output of ten tweets during the week (after week 2’s solitary effort), can only provide some very limited consolation for the LNP, even if we ignore the fact that eight of those tweets were retweets: even in combination, the 69 tweets posted by those four LNP accounts during week 3 don’t even come close to matching the 161 tweets posted by @TheQldPremier alone. At least on Twitter, for whatever that’s worth, and at least as far as the task of getting the message out is concerned, we’ve got to call this week for the ALP, by a wide margin.

The weekly activity and visibility tables also bear this out. Only one LNP-run Twitter account appears in the top 20 most active accounts during week 3 – and remarkably, it’s none of the major actors, but controversial candidate @gavking, who continues to be dogged by his rape comments. Otherwise, the list remains dominated by Labor and Greens accounts, with a solitary candidate for Katter’s Australian Party also appearing. (Update: as with the previous two weeks, I initially turned the ALP’s @ryan4glasshouse into a Greens account – fixed now.)

But the comparative lack of activity from LNP accounts doesn’t mean that Twitter users aren’t talking at or about them: the leaderboard of the most @replied and retweeted accounts in our dataset once again shows a close-run race between @Campbell_Newman and @TheQldPremier, and a large gap to the rest of the field. The presence of Newman’s short-lived @CD_Track account is especially notable here: despite its brief moment as a going concern, a substantial number of users nonetheless paid attention to it.

It’s worth noting in this context, by the way, that the vast majority of @mentions of politicians’ accounts are genuine @replies, rather than (manual) retweets: only 16 of @Campbell_Newman’s 1908 @mentions are retweets; only 104 of the 1852 @mentions received by @TheQldPremier. (And as always, we’re only able to track ‘manual’ retweets – “RT @user …” and variations thereof –, not ‘button’ retweets; there’s no real reason that the amount of button and manual retweets should differ substantially, though.) While some portion of those @replies may only be mentions of those politicians, rather than attempts to communicate directly with them, it is nonetheless notable that of the major politicians we’re mainly focussing on, only @TheQldPremier Bligh makes a sustained attempt to respond (via @replies) at least to some of them.

Finally, another network map of the interactions between politicians’ accounts:

Given the overall dominance of ALP accounts this week, it’s not surprising that this network, too, shows ALP accounts in leading positions. The LNP’s @TimNichollsMP does reappear as an active combatant here (mostly directing his tweets at Deputy Premier @AndrewFraserMP), and on the Labor side, it’s @Matt4SurfersALP who most actively takes the fight to the opposition (in the form of @Campbell_Newman and @LNPQLD). Beyond that, we’re mainly seeing the usual intra-party @mentions – especially of generic party accounts @QLDLabor and @LNPQLD mentioning their respective candidates for the Premiership.

We’ll see what the next weeks bring – only another fortnight to go!

First off, a reminder: what I’ve done in week 1, and will continue to do here, is to look not at the #qldvotes or #qldpol hashtags (those are interesting too, but we’ll leave them for another time), but at tweets from and two all the candidate accounts we’ve been able to identify so far (with some help from the Courier-Mail, which maintains a useful Twitter list of almost all accounts). So, what’s included in our dataset are all tweets which were either sent by those accounts, or @mention them (in the form of @replies or manual retweets). In addition to the politicians themselves, we’re also including a few party accounts, as well as the account of former Premier Peter Beattie.

Let’s start again with a look at the principals: the major parties’ generic accounts, the major combatants (whom I profiled here), and Peter Beattie. Compared to week 1, the @QLDLabor party account is the newcomer in this mix: we’ve only been tracking it since 26 Feb., but this also means we do have a full weeks’ worth of data for week 2. Sadly, though, there’s nothing much to be said about it here – with 34 tweets in seven days, it has remained relatively quiet. By comparison, @TheQldPremier Anna Bligh continues to lead the pack, maintaining her work rate at levels comparable to last week (160 tweets in week 2, just over 180 in week 1). Her deputy @AndrewFraserMP has substantially increased his tweeting, on the other hand: he’s been tweeting roughly twice as much as he did during week 1. By contrast, crucial Labor MP for Ashgrove @katejonesmp, who is in a direct electoral contest against would-be LNP Premier Campbell Newman, has dropped back slightly: posting just over 60 tweets in week 1, she’s only managed 41 in week 2.

The same goes for Peter Beattie (@SmartState1): he’s posted some 60 fewer tweets than in week 1. This is perhaps unsurprising: now working as an election commentator, the start of the election campaign proper would have afforded him a much greater opportunity for comments than the (in Queensland, at least) relatively uneventful second week of the campaign.

On the other side of politics, party account @LNPQLD has more or less maintained a high level of activity: from just under 110 tweets in week 1, its work rate has increased to touch on the 120 mark in week 2. It remains evident that the Twitter campaign here is centred on the party headquarters: where ALP candidates well out-tweet its party account, the situation is reversed for the LNP. With 34 tweets in week 2, candidate for Premier @Campbell_Newman matched his week 1 rate almost exactly, while nominal parliamentary opposition leader Jeff Seeney remains absent from Twitter, and his deputy @TimNichollsMP might as well be: he tweeted just once during week 2.

What is also interesting in examining these week 2 numbers is how much the tweeting styles of these accounts have remained the same from one week to the other (see the corresponding graph from week 1 for comparison). Bligh and her predecessor Beattie remain highly conversational, posting a substantial number of @replies and no retweets (with Bligh posting a significantly larger number of original tweets – i.e. neither @reply nor retweet – and tweets containing URLs); Newman’s tweets follow a similar pattern, if on a much smaller scale. Fraser and the LNP party account both post an almost even mix of original tweets, retweets, and @replies (for both, the retweet component has grown in week 2), but @LNPQLD’s tweets contain more URLs. Jones’s activities are the only where some change is notable: she has taken to posting more retweets in week 2.

Overall activity levels in interactions with these accounts have remained more or less stable during week 2, compared to the levels established by the end of week 1 (on average, there were some 1200 tweets from or to the accounts we’re tracking here). I’ll skip the relevant graph in this post, and will include one which shows patterns since late January in my week 3 post.

Things become more interesting again when we look at the most active and most visible accounts during week 2. Here are the relevant tables, again colour-coded by party:

 
Compared again to last week’s tables, a few patterns are obvious. Some Greens accounts continue to campaign strongly via Twitter; a third second account, the unfortunately-named @GREENS4GlssHse, has joined the lead group (I initially mistook the ALP’s @ryan4glasshouse for a Greens account, too – fixed now). Similarly, three more ALP Twitter accounts have entered the top 20 list in week 2, in addition to @TheQldPremier herself. Overall, and notwithstanding the strong performance of the central @LNPQLD account, this continues to give the impression that candidates from left-of-centre parties in Australia have taken to Twitter with significantly more enthusiasm than conservatives (and interestingly, the three Katter’s Australia Party accounts which were in the top 20 in week 1 have now disappeared from the list). Of the non-official accounts, @Qlder continues to tweet most actively at the accounts we’re tracking here.

But activity does not equal visibility or success: when we examine the @mentions of major political accounts, the picture is considerably more balanced. In terms of @mentions, in week 2 challenger @Campbell_Newman has (just) taken the lead from @TheQldPremier, and these two accounts each receive easily more than twice as many @mentions than any other users listed here. Amongst those other accounts, there are few surprises: Fraser, Jones, Beattie, and the @LNPQLD account all maintain their positions in the lead of the also-rans; @gavking, the LNP candidate under pressure for some highly controversial remarks about rape, also continues to receive a substantial number of @mentions.

The minor parties (in particular, the Greens and KAP) remain comparatively absent from the list, on the other hand; the focus of the discussion, at least to the extent that it is expressed through tweets which directly mention politicians’ accounts (which is what our dataset represents, you’ll remember), clearly is on the major parties – and of these, mainly on their two candidates for the Premiership. Outside of this, there’s a smattering of news-related accounts (in addition to Beattie’s @SmartState1, accounts for Network Ten, the Courier-Mail, and the Brisbane Times), as well as the anonymous account of LNP critic @LNPInsider and the account of Young LNP president Ben Riley (@YLNPPresident; not coloured in the table above since he’s not a candidate in the election, as far as I know).

Finally (for week 2), another quick graph of the network of @mentions between the major accounts (click to enlarge). Here, I’m again showing the political accounts (plus Peter Beattie) only, and compared to last week, I’ve coloured the arrows between them by the party affiliation of the sender (so, a red arrow shows @mentions by an ALP account, a blue arrow shows @mentions by an LNP account, etc.). Node size indicates the number of @mentions received – but only from the accounts shown here (so @TheQldPremier is larger than @Campbell_Newman in this graph, since – due to the greater number of active ALP accounts on Twitter – she receives more mentions from political accounts than he does).

There’s now a fair amount of separation visible between the two camps. Labor accounts mainly tweet at each other (and most of all, @mention the Premier); LNP accounts, to the extent that they’re particularly active at all, do the same. The major exceptions are @AndrewFraserMP and @Matt4SurfersALP on the ALP side, and @ScottDriscollAu for the LNP; beyond that, there are also a handful of one-on-one battles (for example between @karenstruthers, ALP, and @DavidGibsonMP, LNP, and between @DavidGibsonMP and @ryan4glasshouse) or unrequited attacks (the LNP’s @jplangbroek’s tweets to the @QLDLabor account, for example, seem to have remained unanswered).

So much for week 2, then. A post on developments in week 3 will be coming up later today.