Patterns of talk on twitter during the queensland3

Editor's Notes

• #2: To put this presentation in context:I worked with a team of researchers to put together a pilot study of Twitter use during the Queensland floods, which has followed on to extensive work on the uses of social media during crisis events. Jean Burgess, Kate Crawford, and Axel Bruns have a Linkage Grant to study these themes, and have presented their research on both the Queensland floods and the New Zealand earthquakes in many places.
• #3: Context for this research.Encourage anyone interested to download the floods report from the CCi website.
• #4: Between December 2010 and January 2011, the Australian state of Queensland experienced major flooding, causing the deaths of over 30 people, as well as widespread damage to infrastructure and property. The most acute period of flooding occurred in early January 2011: on 10 January, the city of Toowoomba and the Lockyer Valley area were devastated by unexpected flash floods following record levels of torrential rain. On the 11th and 12th of January, large areas of the capital city Brisbane and neighbouring Ipswich were inundated by the rapidly rising Brisbane river, affecting thousands of homes.At the time –shortly after the floods – my role was to code and analyse the raw data from the Queensland floods hashtag and draw out a number of significant themes, and this is what I’m going to talk about today - the themes of talk on Twitter during the Queensland floods, as well as some of our key findings overall.
• #5: we extracted all tweets written by or addressed to the @QPSMedia account within the #qldfloodshashtag. We also developed a representative sample of tweets from the overall #qldfloods data, by capturing every 20th tweet. This enabled us to compare the communicative practices that participants directed toward the emergency authority, as compared to communication styles used in the overall discussion. In previous papers the team has focused on the features of communication with and by the @QPSMedia account predominantly. In this discussion however I’d like to do something a bit different, focusing instead on the communicative styles in the overall #qldfloodshashtag sample. I will provide some examples of each style of communication and the kinds of talk about disasters that participants engaged in in this overall sample, as well as how these communication styles were distributedThen I will briefly introduce a number of concepts and themes that emerged from our research. My hope is to be able to discuss these themes and their relation to other work in the field, through the research represented in this workshop.
• #6: This is an hour by hour visualisation of the peakvolume of tweets within the #qldfloodstweetstream. It coincided with the peak days of flooding in the most populated areas of Brisbane and Ipswich, as well as the start of the recovery period in those areas.
• #7: Drawing from the crisis communications literature we developed a set of thematic and purposive categories and applied these to both the @QPSMedia sample and the baseline sample from the general #qldfloods tweet stream. We tracked the relative prevalence of these categories over the primary phases of the disaster, calculated the resonance (or retweetrate) of each communication category, and combined these data with a close analysis of tweet content to track shifts in communication patterns.
• #8: This graph shows a breakdown of the communication types in each data set. The left shows the @QPSmedia breakdown, with INFORMATION tweets dominating. The right shows the overall #qldfloodshashtag, with a much more even distribution of communication types. In particular we can see the increased dominance of DISCUSSION AND REACTION type tweets, as well as HELP AND FUNDRAISING. Also narratives of DIRECT EXPERIENCE, shown in blue, become part of the discussion, where it is almost entirely left out of interactions with the @QPSMedia authority.This presentation will focus on DISCUSSION AND REACTION and DIRECT EXPERIENCE type tweets, because they were so comparatively dominant in the broader conversation.
• #9: To show this in a different format, this is how information types evolved over time. This is the @QPSmedia sample, where information really dominates.
• #10: And this is the overall sample, where discussion and reactions dominate, although information remains important particularly in the first peak day.
• #11: Breaking the DISCUSSION AND REACTION category down into its types.The above graph breaks this "DISCUSSION/REACTION" category down into specifics. In the overall #qldfloods sample, discussion tweets were much more likely to be "Personal Reactions" to the emergency - expressions of shock, horror, or amazement. We can also see that the "Thanks" category is approximately equal across both samples. However for @QPSmedia, the majority of gratitude was directed towards @QPSmedia specifically, whereas in the overall sample it was directed toward other actors such as Premier Bligh, and rescue workers, in addition to @QPSmedia and other online official sources.
• #12: Through conversations about the floods on Twitter, participants engaged in the negotiation and framing of how the floods should be understood. Adjunctive discussion refers to the practice of using the crisis to explore other issues, and to frame those issues in relation to the crisis. For example, participants evaluated the performance of politicians in relation to the floods, by saying things like “I have a feeling Kevin Rudd would be more human in the face of the floods than Gillard is” and showing approval and appreciation for Premier Anna Bligh’s role in the crisis. Also, about 10 days after the peak of the floods, the hashtag was used to negotiate an appropriate response to the flood levy, with the majority of participants giving reasons why the levy should be accepted, and in particular trying to correct false assumptions about the levy, such as that it covers private individual flood losses or is a replacement for insurance. One such tweet quoted another tweet against the flood levy, calling the attitude “appallingly selfish”.
• #13: Thanks and gratitude type tweets were really used to negotiate how the floods should be remembered, with an overwhelming amount of gratitude, as I said earlier, directed toward Anna Bligh. This wordle shows the words used in these tweets.
• #14: Another subset of tweets were the meta-discussion tweets, where people talked on twitter about why twitter was so important during the crisis. They gave more than one reason for this. One reason, shown on the right, was that they felt that social media was important for crisis communication, sharing of situational information, and sharing information on how to help. Certainly these claims are borne out by our research on the Queensland Police media account, which was able to disseminate information quickly and accurately, and was particularly useful for mythbusting IN-accurate information. The other theme of meta tweets were the sense that participants were part of something important or necessary. They talked about the affective importance of “being on Twitter” and about being a “retweet bot” for the good of the floods. Such tweets coincided with our findings in the @QPSmedia report that people were much more likely to retweet during the crisis, with a particular focus on re-tweeting with the effect of amplifying informational tweets. People expressed excitement and satisfaction about the important role THEY were also playing in this crisis, by expressing the sense of importance with which they viewed their amplification and sharing role.
• #15: For non-local participants, messages of support were the biggest contribution to the #qldfloodstweetstream. These seemed to follow a very similar format and contained particular kinds of language.
• #17: This category of tweet was unique to the #qldfloods stream not represented in the @QPSMedia sample. There was a sense in which something is happening – people were narrating their experiences and describing their part in the floods in community with others. Because participants in the #qldfloodstweetstream included both local people and non-local people, it was used to share and describe experiences of what it was like to be on the ground. People talked about stocking up their pantries and preparing for evacuation. In particular the narration of the clean up and recovery process was a strong theme in these kinds of tweets. People were expressing being part of the flood clean-up through their Twitter accounts, addressing both their own followers and the spontaneous network of twitter users around the #qldfloodshashtag.
• #18: Another use of the personal narrative type tweet was sending updates to twitter followers to reassure them of their safety and of the safety of their family members. A lot of people tweeted things like “I’m okay” and “I called my family and they’re okay”. Since these tweets went out not only to their own followers but also to the broader Queensland floods network, it was another way of expressing involvement and entanglement in the crisis. These findings relate to the concept of “digital convergence” from the work of
• #19: In the literature on uses of social media during crisis events, Goolsby describes how ad-hoc crisis communities tend to form around emergencies, and observes that in these cases Twitter becomes a crisis platform by accident rather than by design. She argues that this results in chaotic and ad-hoc strategies for coping with emergencies, such as the retweeting of situational information. Hermida argues that crisis news becomes a social experience – a distributed conversation in which people respond collectively to crises. Close analysis of each communicative genre makes clear how people felt about using social media in this way, how they benefited from it and appreciated it during the time of crisis.Liu argues that social media users participate in practices of “socially distributed curation” and collective memory negotiation during crisis events”. Robinson found that social media users actively negotiate and construct the meaning of particular events.
• #20: Digital convergence relates to the concept of “convergence behaviour” from studies of crisis sociology. Convergence behaviour refers to the tendencies of individuals during crisis events to converge on the site of the crisis, practising behaviours such as site-seeing or pilgrimage. Crisis informatics scholars extend the concept to describe how non-local individuals converge around emergencies in online contexts, to show support and satisfy curiosity. Finally we found that people engaged in behaviours of amplification, retweeting information they considered important for others in order to be part of the process of crisis response and information sharing. @QPSmedia would not have been as successful as they were without these behaviours of amplification. People’s sense of involvement is also shown in the meta tweets, where they say things like “it’s really important to be on Twitter today”.
• #21: I am interested in hearing from people at this workshop any intersections with your own work – not only in terms of other teams looking at social media use during crisis events – but also the categories of talk that go on during crises. The sharing of experiences, the framing of political actors, and the policing of others’ behaviour as a feature of ordinary people’s responses to crisis events.