Image: "Multiple Tweets Plain" by

The global interconnected network of computers known as the Web offers opportunities  for new forms of social analysis – not least the possibilities offered by ‘big data’  – digital data sets so large they require novel computer  processing and software. These new kinds of data provide the possibility of real time analysis that offer the potential to find out what is happening ‘right now’ on a population level scale, whilst also minimising analytical costs. Another, related aspect of big data is the social media element. One area of the Web that has already received considerable research attention is Twitter. With an estimated 200 million users creating over 400 million Tweets it has become a vast repository of user generated textual data.

Unfortunately much ‘research’ about Twitter can be characterised as ‘Twitterology’ a practice that rates somewhere below astrology in terms of its ability to deliver meaningful interpretations of the social world. This includes simplistic content analysis  to engage in sentiment analysis by collating vast numbers of tweets and counting keywords – like ‘happy’ – to make claims such as “Twitter Proves the West Coast is Happier”.  Surely sociologists have at our fingertips a set of analytical methods that can complement technical forays into the Web to make much more meaningful claims about social life than this?

We have developed computer software to dynamically model Tweets to explore the network properties of this part of the Web. Using this we can follow actors and interlinked chains of Tweets, identifying initiators and amplifiers, and exploring twitter data flows over time. This produces some interesting patterns, as our example below demonstrates. We harvested 677 tweets that used the hashtag #saveournhs and followed the network  of interactions around this tag. In the video the earliest tweets migrate to the right of the display, displaced by a busier larger network on the left.  At its heart is one user (@UnitetheUnion) in a circle or node which quickly turns red, signalling its importance to the network as a highly retweeted source.  As the model runs @UnitetheUnion remain dominant; but interestingly this is on the basis of just four Tweets on the hashtag #saveourNHS.  These four messages propagate through the network extremely rapidly and produce visibly  longer chains of retweets.

On the right of the display another interesting interaction emerges. First a node called @999CallforNHS, which originated at the start of the model, links temporarily to the network surrounding @UnitetheUnion then it breaks off before re-joining the network some 0.43 minutes into the display. The linking node here which also becomes significant (turning red at 1.28) is a user ‘called’ @CraigFarlow. The latter, at the time of this analysis had penned just 812 tweets, was following 65 people and being followed by 28. By contrast @999CallforNHS had 1487 followers, followed 1169 and had made 3210 tweets. These two nodes made just 35 and 25 tweets using the #SaveourNHS tag and received 77 and 51 retweets respectively.  While none of their tweets produced a retweet chain as long as @UnitetheUnion their activity secured their places as significant (red) nodes in the network.

The yellow nodes which emerge in the centre of the display at around 0.50 minutes in the video are also interesting. One of these is @NHAparty. This node is the source of two of the longest retweet chains and provides another bridge between the two networks. The NHS Action Party – represented by this ‘user’ is a new political party that was formed by two doctors, Richard Taylor and Clive Peedell, in 2012 to “fight for the original ideals of the NHS” . This node’s position, is suggestive of what has been referred to  as ‘bridging capital’ and acts as a kind of “sociological WD-40” lubricating the connections between otherwise disparate actors.

Because of our own roles as citizens and researchers in UK health care we can make more sense of these Twitter analytics by drawing on other knowledge and data about the networks and the actors.  For example we can confirm that @UnitetheUnion is a social media account for a large trades union in the UK. We can see on Twitter that it has 21688 followers and has authored 7000 tweets. We might through further digging find out that the Union has 1.4 million members, and learn more about their structural organisation and their role in the event on 29 September . We can take an educated guess that its Tweets are produced by a media team, possibly with skills in press relations and media communication and that this node therefore represents an actor that is a bit different to other Tweeters who might be lone individuals.

In our work we have been exploring the potential of mixed methods for analysing Twitter and have called for a shift from big to wide data(1). Wide data might include network and content data augmented with additional data sources and analysed from an interpretive perspective. We could collect political speeches, news media and other web coverage to provide further details about the NHS Action party, and Drs Peedel and Taylor.  We might remember that Taylor ‘has form’ as an activist for the NHS, having stood as an independent MP on a single health issue (fighting the closure of Kidderminster hospital in 2001). If we had time, and really wanted to broaden our analysis and make or data truly wide, we might have chosen to attend the event on 29/9 or attempted to contact some of the actors who did, or we could interview the authors of the tweets to understand more about their motivations and subjective experiences of the interaction.

By augmenting Twitter analytics with additional data we could make more sense of the network and perhaps begin to understand the wider interactions, political discourses and events that they represent or allude to.  This at the very least would tell us more than counting emotions in the Twitterverse.

Beyond this, perhaps by developing a more nuanced understanding of networks of resistance as they play out on new social media we might,  as concerned citizens,  and perhaps as activists ourselves, be able to use our sociological insights to support or inform activism (and maybe to help Save Our NHS).

About the Authors: Catherine Pope (@cj_pope) is Professor of Medical Sociology, and a former Deputy director the Southampton Web Science Doctoral Training Centre, and she was fortunate to be one of Ramine’s co-supervisors (with Susan Halford and Les Carr) and so has spent the past three years vicariously learning about Computer Science. Ramine Tinati (@raminetinati) has recently completed his PhD in Web Science at the University of Southampton. Originally trained in Computer Science he has spent the past 3 years exploring how Actor Network Theory might be applied to the Web and open government data