{"id":690,"date":"2011-04-05T14:00:00","date_gmt":"2011-04-05T04:00:00","guid":{"rendered":"http:\/\/www.mappingonlinepublics.net\/dev\/2011\/04\/05\/broader-twitter-patterns-during-acute-events\/"},"modified":"2012-04-10T13:56:02","modified_gmt":"2012-04-10T03:56:02","slug":"broader-twitter-patterns-during-acute-events","status":"publish","type":"post","link":"https:\/\/mappingonlinepublics.net\/dev\/2011\/04\/05\/broader-twitter-patterns-during-acute-events\/","title":{"rendered":"Broader Twitter Patterns during Acute Events"},"content":{"rendered":"

Working through our available data on Twitter<\/em> use during crisis events ahead of the Eidos Institute symposium on Monday, I started thinking about some of the broader patterns we are seeing. Very obviously, a good bit of the #hashtag activity around acute events is taken up with retweeting information – both simply passing it along unedited, and adding further details or commentary in edited, manual retweets. Additionally, there is a certain amount of @replying between participants, though such follow-on conversations tend not to include the #hashtag much any more, unless the conversants deliberately seek to make their discussion visible to the wider #hashtag community (to perform it publicly, in other words).<\/p>\n

So, the question becomes: how much of the #hashtag space around specific acute events is taken up by these forms, and how much consists of new tweets that are neither retweets nor @replies? That’s what I want to explore in this post, for the key examples of the Queensland floods (#qldfloods), the Christchurch earthquake (#eqnz), and the Japanese tsunami (#tsunami). I also need to note again that our data on retweets includes only <\/em>manual, old-style retweets (‘RT @[user]’), not retweets made using the Twitter<\/em> retweet button – but that’s actually a valuable limitation in our current context, since a manual retweet by definition requires users to make a more conscious effort than a mere press of the retweet button.<\/p>\n

<\/p>\n

First, then, here’s the picture for #qldfloods, for the core days from 10 January 2011, when the floods hit Toowoomba and the Lockyer Valley, to 16 January, when the immediate flood crisis in Brisbane had passed (with all of these graphs, click to enlarge; all dates and times in AEST):<\/p>\n

\"image\"<\/a> <\/p>\n

Shown here is the total volume of #qldfloods tweets (in blue in the background, value axis on the left) and the percentage of those tweets which were manual retweets (in red) and @replies (in green, axis on the right). Overall, over the course of these days, #qldfloods consists of 52% retweets and 14% @replies – leaving 34% non-retweet\/@reply tweets – and these percentages are relatively steady: there’s only a very slight decline in the average retweet percentage, and a similarly slight increase in the proportion of @replies.<\/p>\n

For #eqnz – tracking the days following the 22 February 2011 earthquake -, the picture is relatively similar:<\/p>\n

\"image\"<\/a> <\/p>\n

The #eqnz hashtag data consists of 55% retweets and 13% @replies – leaving 32% non-retweet\/@reply tweets. Here, however (and I’ve indicated this by inserting a linear trendline) the average retweet percentage drops from just over 60% to just under 50% during the week following the disaster – an indication, perhaps, of the shift from an ‘ambient journalism’ phase to a ‘recovery’ phase in the use of Twitter as I’ve outlined it in my previous posts on #eqnz<\/a>.<\/p>\n

Our third example, Twitter<\/em> activity around the Japanese tsunami, comes in two forms. Because of the prevalence of the term ‘tsunami’ as both keyword and #hashtag, I was able to capture all tweets containing the word itself, and so we’re able to examine both the greater number of tweets which simply use the term ‘tsunami’, and the smaller subset of those tweets which deliberately include #tsunami as a hashtag. Starting with the hashtag, over the days following the initial earthquake on 11 March 2011:<\/p>\n

\"image\"<\/a> <\/p>\n

Here, the patterns diverge very strongly from our previous two examples: on average, we see around 65% retweets, and only 7% @replies – and the trendlines are virtually flat for retweets (or even very<\/em> slightly pointing upwards), and rising from 6 to 9% for @replies.<\/p>\n

But perhaps that’s not surprising: the Japanese earthquake and tsunami constituted a much longer disaster event than the Queensland floods or Christchurch earthquake – indeed, with the current nuclear crisis, they’re arguably still continuing -, and the destruction they caused also surpassed the other two crisis by an order of magnitudes (which also resulted in even more intense and prolonged attention from Twitter<\/em> users). Given the massive global audience for news from Japan, we should<\/em> expect a higher number of retweets of news reports and other information, and that’s precisely what we’re seeing here. <\/p>\n

Indeed, while I’m not prepared to base this theory on only three data points right now, it’s an intriguing thought that the volume of retweets related to an event may provide a relatively accurate measure of the global attention to it: so far, we’re certainly seeing a pattern which would suit that reading – from 52% and 55% retweets in #qldfloods and #eqnz, which were perhaps roughly comparable in global media impact (though I should stress that this is by no means meant as a comparison of which was more devastating to the local population), to 65% in #tsunami. We’ll have to track future events (also well beyond natural disasters) to substantiate this theory, at any rate.<\/p>\n

Looking at tweets containing the keyword<\/em> ‘tsunami’ (which includes the #tsunami hashtag tweets, of course, but also many more tweets beyond them), a different picture emerges:<\/p>\n

\"image\"<\/a> <\/p>\n

Of this larger collection of tweets, a smaller number are retweets, while there are more @replies: the overall retweet percentage is 49%, and the @reply percentage is 11%. That’s hardly surprising, given what I’ve already said about #hashtags as a means of coordinating a conversation: @replies which mention<\/em> the word ‘tsunami’ aren’t necessarily also hashtagged<\/em> #tsunami, so we see a higher percentage of @replies here; conversely, hashtagged tweets about the tsunami are probably more likely to be seen and retweeted than non-hashtagged tweets that simply contain the word ‘tsunami’, resulting in a much lower retweet percentage for tsunami keyword tweets compared to #tsunami hashtag tweets. <\/p>\n

Additionally, there are also notable trends in the data: a decline in the retweet percentage, from around 53 to 45% (which, given that we’ve seen above that the percentage of #tsunami retweets is stable, must be driven solely by a decline in the proportion of non-hashtagged ‘tsunami’ retweets), and a rise in @replies, roughly from 10 to 13%.<\/p>\n

Again, it’s too early to proclaim any grand patterns in disaster response tweeting from this limited number of cases, of course – but further exploration of additional cases may point us to clearer conclusions. (The specific context of each case must be kept in mind in doing so, however – as I’ve noted, the tsunami and its continuing aftermath are a very different case from the comparatively more contained, if similarly devastating, disasters in Queensland and Christchurch.) <\/p>\n

When I began the process of looking at the data from this perspective, my hypothesis was that retweeting would decline as a percentage of the total number of tweets once the immediate (global) attention to the event passed and the #hashtag communities were driven more strongly again by directly affected locals sharing information and organising their activities, rather than by a global audience retweeting news stories and first-hand footage; my results so far are too mixed to confirm that hypothesis at this point, however. <\/p>\n

The #eqnz patterns clearly fit my assumption; the #qldfloods retweet percentage remains steady, though, and does not decline over time (but the decline could be concealed in part by discussion of other, unrelated flood events in Queensland, or by a local discussion about the value of the #qldfloods hashtag itself which immediately followed the emergency). The #tsunami retweet percentage also remains flat (for the hashtag), but given the longer duration of this continuing series of disasters, any decline in numbers simply might not have started yet during the five days we’re looking at here; we’ll have to look again at a later date. Conversely, the ‘tsunami’ keyword<\/em> data does show a decline in the percentage of retweets, but given the complexities of this event, I’m not willing to put too much stock in this so far.<\/p>\n

So, I’m not quite sure what exactly to make of all of this so far, but I’m keen to explore it further as we gather more comparable data. Any comments and feedback welcome!<\/p>\n","protected":false},"excerpt":{"rendered":"

Working through our available data on Twitter use during crisis events ahead of the Eidos Institute symposium on Monday, I started thinking about some of the broader patterns we are seeing. Very obviously, a good bit of the #hashtag activity around acute events is taken up with retweeting information – both simply passing it along … <\/p>\n

Continue reading “Broader Twitter Patterns during Acute Events”<\/span><\/a><\/p>\n

<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[180,172,8,177],"tags":[92,73,99,74,100,298],"class_list":["post-690","post","type-post","status-publish","format-standard","hentry","category-analysis","category-crisis-2","category-twitter","category-visualisation","tag-eqnz","tag-qldfloods","tag-tsunami","tag-disaster","tag-retweets","tag-twitter","entry"],"_links":{"self":[{"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/posts\/690","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/comments?post=690"}],"version-history":[{"count":0,"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/posts\/690\/revisions"}],"wp:attachment":[{"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/media?parent=690"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/categories?post=690"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mappingonlinepublics.net\/dev\/wp-json\/wp\/v2\/tags?post=690"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}