Blogs

The final paper in this AoIR 2013 crisis communication panel is by Jean Burgess, Avijit Paul and me. The slides are below, and audio will follow later is now online as well...

Cumulative and Comparative Social Media Analytics for Crisis Communication from Jean Burgess

The third speaker on our AoIR 2013 crisis communication panel is Megan Finn. She begins by noting that the US Geological Survey is now using Twitter data to detect earthquakes - but more generally, there are also limits to the use of Twitter and other social media data, as not all groups in society are equally represented in such data, and in social media as such.

A disaster is traditionally defined as an event, concentrated in time and space, in which society undergoes severe data and essential functions of society are interrupted. But the components of this definition are problematic - a crisis is more a reflection of the ability or otherwise of the socioeconomic system to cope with unusual conditions in a current situation, and this needs to be better recognised; for one, the recovery period also emerges as an important point of focus here.

The next speaker in our AoIR 2013 panel on crisis communication is Andres Monroy-Hernandes, who focusses on emergency responses in the current Mexican drug war. Traditionally, emergency information has been disseminated by government officials and the media, but this is not necessarily the case in Mexico, due to the scale of civil disorder in the country: journalists and government organisations in northern Mexico are essentially operating under a self-imposed news blackout due to the pressure they feel from the druglords.

Instead, social media are increasingly adopted for information: citizens in lawless areas are warning each other of "risky situations" (shootings, bombs, etc.), with hashtags like #mtyfollow emerging as the mechanisms to collate such warnings. A kind of "narco language" is also emerging - for example for kidnappings, dead bodies, etc. - and the occurrence of such language is correlated with the murder rate in specific areas, and with the magnitude of specific events.

The next panel at AoIR 2013 is on crisis communication, and we have a paper in this one, too... We start, however, with Leysia Palen from the fabulous Project EPIC in Boulder, who begins with a general overview. Disasters are disruptive, unpredicted events which mean that normal daily routines cannot continue; emergencies become disasters when they overtax available local resources.

One aspect of disasters is mass convergence: a slower-motion convergence of people either in local locations or in spaces immediately outside the disaster zone - including affected residents, support staff, and curious onlookers. These groups are often organised around available items of information about the disaster.

The final speaker in this AoIR 2013 panel is Tim Highfield, whose focus is on the doping scandal surrounding Lance Armstrong between August 2013 and January 2013, from USADA stripping Armstrong of his titles to his confession interview with Oprah. This draws on a number of different datasets, including tweets mentioning @lancearmstrong during the 2012 Tour de France as well as tweets mentioning 'Armstrong' in subsequent months.

During the 2012 Tour, reports emerged that several riders had testified against Armstrong, generating substantial discussion with the Tour de France Twitter community as an extension of the fan/athlete para social interaction - with fans making both critical and supportive statements towards Armstrong. Subsequent spikes in activity surround Armstrong's stepping down from the Livestrong foundation and the Oprah interviews.

The second presenter in our AoIR 2013 celebrity crisis panel is Ana Vimieiro, whose focus is on the crisis around Oscar Pistorius following the death of his girlfriend Reeva Steenkamp. Ana's approach in this is to use frame analysis to explore how the crisis is being conceptualised by the users discussing it on Twitter.

There are various approaches to frame analysis: holistic variable coding, a textual focus on terms and expressions (which ultimately focusses on topics more than frames), and data reduction techniques which draw on clustering approaches. Ana is employing the latter model, coding the text for less subjective categories including sources, causes, solutions, consequences, metaphors, slogans, examples, and moral judgments.

The post-lunch session at AoIR 2013 starts with a panel on celebrity crises, which has now become a QUT-only affair. We're starting with a paper by Theresa Sauter and me, on the Pope's @pontifex account. (Slides and audio are below.) Celebrity accounts in general are one of the big drivers of Twitter activity, as Twitter itself positions them. The @pontifex account was set up in December 2012 by Benedict XVI, with nine different language accounts set up, including one in Latin.

#Pontiff-Ex: The Twitter Community’s Reaction to the Papal Resignation from Axel Bruns

We tracked the English-language account, which had gathered 1.6 million followers by the time Benedict XVI resigned. He'd mainly posted brief prayer-style messages, coordinated across all languages, and the new pope Francis has done much the same since then. This is perhaps unsurprising given the status of the Pope.

The final speaker in our AoIR 2013 panel is Fabio Giglietto, who presents a season-long study of Twitter use alongside 11 Italian TV talkshows. Twitter use alongside such shows can reveal the power struggles between political and media actors and everyday citizens.

Fabio's team bought Gnip data for the relevant hashtags related to these 11 talkshows, adding up to some 2.5 million shows. 76% of these were made while the shows were on air, with some 187,000 unique on-air contributors. The team also identified the key peaks in engagement, and associated them with specific time windows within the broadcasts; they then conducted a content analysis of the tweets posted and the content broadcast during these windows.