Scottish referendum 2014: Twitter as a campaigning medium
The organisation of the competing spaces of the ‘yes’ and ‘no’ campaign on Twitter duringthe campaigning period Master’s thesis
Marissa Sieuwerts, MA New Media and Digital Culture
marissa.sieuwerts@gmail.com
Dr. S. Milan
University of Amsterdam Final version: 26 June 2015
CONTENTS
0. ABSTRACT 3
1. INTRODUCTION ... 4
1.1 BRIEF OUTLINE OF RESEARCH ... 4
1.2 BACKGROUND OF THE DEBATE ... 4
2. LITERATURE REVIEW ... 6
2.1 CLASSIC APPROACHES TOWARDS POLITICAL COMMUNICATION -‐ ... 7
2.1.1 Politics on the web and web campaigning: an introduction ... 8
2.1.2 Social media and political communication ... 10
2.2 CONTEXTUALISATION: TOWARDS A MULTI-‐DISCIPLINARY APPROACH ... 12
2.2.1 Demand for shift of paradigm ... 12
2.3 MEDIUM SPECIFICITY: TWITTER AND THE LOGIC OF PLATFORMS ... 14
2.4 APPROACHES TO TWITTER STUDIES: DIGITAL METHODS ... 16
2.5 APPROACHES TO TWITTER STUDIES: SIMILAR CASE STUDIES ... 17
2.6 NETWORKS AND TOPOLOGICAL THEORY ... 20
2.7 SOCIAL METRICS ... 23
2.8 BIG DATA AND ANALYTICS ... 24
3. METHOD ... 26
3.1 SUB TOPICS ... 26
3.2 DATA COLLECTION ... 27
3.2.1 Hashtag analysis (#) ... 29
3.2.2 Conversation network analysis (@mentions) ... 29
3.2.3 Retweet analysis (RT) ... 30
3.3 DATA VISUALISATION ... 31
4. FINDINGS ... 32
4.1.1 Co-‐hashtag analysis ... 36
4.2 CONVERSATION NETWORK ANALYSIS (@MENTIONS) ... 40
4.2.1 Bi-‐partite hashtag-‐mention graph ... 43
4.3 RETWEETS ANALYSIS (RT) ... 47
4.3.1 Retweets: Network effects ... 50
5. CONCLUSIONS ... 56
6. DISCUSSION ... 60
6.1 SOCIAL METRICS: TOP RESULTS ... 60
6.2 IMPLICATIONS OF USED METHOD ... 62
6.3 OPPORTUNITIES FOR FURTHER RESEARCH ... 64
7. REFERENCES ... 65 8. APPENDIX ... 69
Cover image by BBC news (http://www.bbc.com/news/blogs-‐trending-‐29268773) last accessed: 23-‐6-‐2015
ABSTRACT
This thesis aims to look at Twitter as a broader political platform showing societal trends and changes, and seeks to look at how competing spaces of a debate, in the case of this study the two competing political camps during the recent Scottish referendum, are being organised through Twitter as a social medium.
By taking the development of a new methodological framework in the field of political communication as one of the main research goals, this paper seeks to assess particular claims as they were made by an early research to the Scottish referendum and the debating sphere on Twitter. This study, undertaken by the University of Glasgow during late 2013, is for this thesis used as a starting point, aiming to critically assess those findings, backing them up with empirical research.
This paper therefore looks at Twitter as a social research platform in the means of a networked content analysis, hereby ‘debanalising’ Twitter by looking at political networks, taking its value for research into account. Above all, it emphasises the development of an updated approach towards studying political communication, an academic field that, as will be argued, could be in need for an update of the research methods.
In the long run, the main research objective is, besides mapping a political debate, to consider existing methods and open up the possibility for methodological discussion within the field of political communication, while developing and proposing a different methodology to study political communication on Twitter.
KEYWORDS
Scottish independence referendum, Twitter, political communication, networked content analysis, political networks, platform studies, digital methods, big data, data visualisation
1. INTRODUCTION
1.1 Brief outline of research
This thesis covers the popular social media microblogging platform Twitter, and more specifically the use of Twitter in political communication. Last year’s independence referendum in Scotland is used as a case study to look at how this political debate and campaigning is being organised on Twitter. The main research objective is, however, besides mapping a political debate, the development of a new methodology through a multi-‐disciplinary approach, by merging the fields of political communication and critical new media studies (more specifically digital methods) together.
Taking an existing study to the topic of the Scottish referendum into account,
undertaken by the University of Glasgow, it was sought to assess the claims made by the researchers. This research looked at how the debate around the referendum on Twitter in organised in an early phase of campaigning, namely the end of 2013. Considering Tweets sent during a week (13-‐12-‐2013 until 19-‐12-‐2013), the study concluded that the ‘yes’ campaign is more active on Twitter and has a greater reach1. Furthermore, they
concluded that debate tends to not to be organised through the ‘official’ campaigning: rather, it is decentralised. Conversations are not channelled through the official campaign2, all of which is visible in their Gephi network graph3.
As these results were not turned in a full academic paper, I aim to assess these claims with empirical research, approaching the topic multidisciplinary. Therefore, whilst studying the debate around the referendum, a methodological framework for studying political communication on Twitter was developed, as will be proposed and critically approached in this paper.
1.2 Background of the debate
The main case study that will be considered in this paper is the Twittersphere around the political Scottish referendum debate. On 18 September 2014, after years of debate
1 http://www.bbc.co.uk/news/uk-‐scotland-‐scotland-‐politics-‐25642809 last accessed 23-‐6-‐2015 2 http://policyscotland.gla.ac.uk/twitter-‐analysis/ retrieved last accessed 23-‐6-‐2015
around leaving the unity of the United Kingdom and becoming an independent nation state, Scotland organised a referendum which concerned the question Should Scotland be an independent country? This question could be answered with either ‘yes’ or ‘no’ by Scottish nationals in the official referendum last year. The main election campaign was focused around the two sides the debate, on the one hand there was Alex Salmond’s ‘Yes Scotland’ campaign4 for becoming an independent country and on the other, Alistair
Darling’s Better Together campaign5, which favoured staying in the union.
When the results of the referendum were announced the next morning, it was the ‘No’ side of the campaign that could claim the victory: the "No" side won, with 2,001,926 (55.3%) votes against independence and 1,617,989 (44.7%) votes in favour, with a record turnout of 84.6% of the people that were eligible to vote6. In this 2014
referendum, 16-‐ and 17-‐year-‐olds were allowed to vote for the first time. The final results per county are visible in the map in figure 1, which shows the majority of counties with a greater ‘no’ vote coloured in shades of red.
Fig 1. In this map7, the final results of the vote are visualised, with the darker red shades corresponding to
a greater ‘no’ vote, while green represents an area voting ‘yes’.
4 http://www.yesscotland.net/ last accessed 23-‐6-‐2015 5 http://www.bettertogether.net/ last accessed 23-‐6-‐2015
6 Data: http://en.wikipedia.org/wiki/Scottish_independence_referendum,_2014 last accessed 23-‐6-‐2015 7 Map: http://en.wikipedia.org/wiki/Scottish_independence_referendum,_2014 last accessed 23-‐6-‐2015
The campaign has been going on until and has been crucial until the very last day, as both the yes or no side of the debate did not gain convincing majority in the polls until the day of voting. Social media platforms as for instance Twitter were widely used to organise the debate and create a network; in total, 2.057.874 tweets concerning the Scottish referendum were collected by the TCAT Twitter scraping tool8 (Twitter
Capturing and Analysis, a project of the Digital Methods Initiative Amsterdam), during the timespan of about a year (21-‐11-‐2013 – 02-‐12-‐2014). This paper focuses on Twitter as a platform and how it organises competing spaces, in this sense two competing political campaigns, whilst developing a multi-‐disciplinary approach towards political themed Twitter studies.
2. LITERATURE REVIEW
Previously, I briefly outlined the multi-‐disciplinary approach of this study with the development of a new methodological framework as one of the main research goals of this thesis: merging methodologies and approaches of both political communication and digital methods as research field. In the following section, I will first introduce the field of political communication and seek to establish an understanding of the classic and current paradigm in the field, outlining its most dominant methodologies, approaches and introduce a variety of significant theories and its contributors to the field.
Furthermore, I will engage with politics on the web and web campaigning, e.g political research in the digital age, before moving on towards the contextualisation of this paper discussing the reasoning for its multi-‐disciplinarily approach. The last part of this
literature review is dedicated to placing the paper into context, by discussing medium specificity and platform studies, and engage with the proposed methodology discussing the dominant research approaches in digital methods, by zooming into case studies with a similar research goal. Lastly, the literature review aims to critically engage with the developed methodology, focusing on critical assessment of its operationalisation by considering networks and topological theory, the debate regarding spoilt data in the sense of social metrics and bias, followed by arguments surrounding the concept ‘Big Data’.
2.1 Classic approaches towards political communication -‐
As a subfield of political science, political communication seems to be a research discipline that is well established as well as accepted and noticed as an academic field that deserves serious attention. Characterised by a variety of approaches and methods to look at specific political communication such as debates, campaigns or advertising, the field offers a resource of research that is properly operationalised. In the following section, I would like to draw the attention towards the ‘classic’ methods, theories, methodologies and ways to undertake research that together constitute and shape the backbone of the field of political communication. However, departing from an
understanding of the classic approaches of the field, I argue that the field might be stuck in old methodologies, hereby creating a homogenous research environment (Nielsen 2014).
First of all, the sample of existing studies that have been considered showed an emphasis on research to televised or press communication; to journalism and mainstream media. Eric Louw (2010), for instance, identifies the process of political communication as a hype-‐making machine that is mainly being organised through mainstream media. Following a constructivist approach, the emergence of the media-‐ tization of politics is mapped; according to Louw, the emerging media-‐attention of journalism and television translated the field of political communication to a new approach; the study of media hypes (Louw 2010).
Since research is often emphasising political communication in mainstream media, studies are generally aiming to look at political advertising or campaigning within the mainstream media landscape. Research is therefore characterised by direct content analysis, an approach that looks at the political message and its contents itself. Aiming to gain insight in the effects of political messages on people, research focuses for instance at how the message affects voting abilities, for instance by seeking to map voting behaviour combined with a qualitative analysis of the content of the political
communication (Kaid and Holz-‐Bacha 2006). A similar approach is leading in the SAGE handbook of political advertising (2006), where the main research goal is mapping (international) televised political communication, specified towards different countries, hence seeking to understand how the process of televised political communication
differs around the world. This approach, which Kaid and Holz-‐Bacha identify as a descriptive rhetorical analysis, looks at narrative, languages and stories present in political advertising, symbols or semiotics, it seeks to describe and make sense of the content using interpretation.
Research in the field is thus characterised by approaches that revolve around cultural effects (representation) and how this is being affected, which is a research objective often being undertaken through the use of surveys. The use of surveys or
interviews in order to measure (cultural) effects of political communication is a common methodology in existing studies in the field. What ultimately matters here is that if the field is generally seeking to undertake research through qualitative approaches, which is not necessarily better or worse; however, solely focusing at qualitative research might contribute to the homogenisation of the research field (Nielsen 2014).
In general, it could be concluded that in the field of political communication, research is focused at, firstly, direct-‐content analysis of the messages themselves in political
campaigning, and secondly, at the effects of political advertising (on voters), or what media does with people, rather than for instance looking at the medium in a political campaign and how this particular medium affects the process of political campaigning, or what people do with media. This could be perceived as a void in the field of political communication, an argument that I will elaborate in section 2.2. The classic
methodologies and key characteristics of the field are featured in a number of existing studies on political communication of the web and web campaigning, which will be discussed in the section that follows.
2.1.1 Politics on the web and web campaigning: an introduction
Looking at political communication that is specifically featured on the web, research tends to be characterised in a similar way as what I previously identified as ‘classic political communication’: studies aiming to unveil how political advertising is being organised, using research that is often undertaken through direct content-‐analysis of the political messages themselves. In the sample of studies that were taken into
consideration in this chapter, research generally focuses on the more static political part of political communication, such as the political messages themselves, rather than the
dynamic communication aspects, such as conversation, user-‐ or voter-‐engagement and debate.
This tendency towards the emphasis of the actual political message is visible in the studies towards web campaigning by Foot and Schneider (2006), while a starting point towards a changing paradigm starts to develop in some of the studies discussed by Baringhorst, Kneip and Niestyo (2009), respectively. Foot and Schneider examine in their research the use of the web in political campaigns in the United States. Through the analysis of websites of US election campaigns over the timespan of 4 years, they aim to trace the development and emergence of web campaigning in the US political
campaigning landscape, and how these practises function to extend and enact with campaign activities (Foot and Schneider 2006). Analysis is based on systematic observation of a large number of campaign websites created by the candidates
themselves, aiming to look at how political campaigns and the web are (becoming more closely) related and are in a constant state of change and evolution (Foot and Schneider 2006). By studying the actual campaigning websites that are involved in the electoral process, seeking to look at the practise of how website visitors are involved in the campaign organisation and how political supporters are informed, connected and mobilised through the content of these websites, this study lies the emphasis on the content layer of political communication on the web, rather than focusing on the contextual layer of media specificity and media attributes (Foot and Schneider 2006).
Now, it might look like that previous studies specifically towards political
communication on the web are solely focused at the political opinions and statements (content) themselves rather than political communicational processes between users and political parties (context). However, a slight change of paradigm, changing the focus of political communication studies from the political message itself more towards the communication processes could be identified in some of the studies that were taken into account by Baringhorst et al. In Political campaigning on the web (2009), the researchers identify ‘evolutionary phases’ of political campaigning and claim that we are currently in the ‘postmodern campaigning’ phase, characterised by a pluralisation of media channels, resulting in ubiquity (Baringhorst et al 2009: 16).
One research in particular, the study by Zielmann and Röttger9 assesses this notion of
postmodern campaigning by looking at characteristics and developments of political parties and campaigns in a variety of countries, focusing at differences between those countries. More specifically, they seek to study if the potentials of web campaigning, which they identify as an extremely emergent aspect of political campaigning, are being fully appropriated and even embraced by the sample set of countries (Baringhorst et al 2009). A proposal towards a change of paradigm is just yet slightly visible here in this study; attention moves from the content of political campaigning to the context of in which it takes place; the digital environment of web campaigning. However, they
conclude, the countries in the sample set still generally characterise web campaigning as ‘an accessory to the offline content of parties’ (Baringhorst et al 2009:71).
2.1.2 Social media and political communication
Besides the research attention towards televised or printed political communication and more general political campaigning websites, previous research towards (online)
political communication and social media has been characterised by a rather specific emphasis on the blogosphere and its most influent bloggers. Drezner and Farrell (2008) for instance, are interested in the blogosphere’s networked role in conveying political statements, and how this might affect political opinion-‐making of its readers and influence voting behaviour or polling. Taking a sample of blogs and analysing them as networked content, Drezner and Farrell argue that the blogosphere is characterised by a rather specific network topology, resulting in the individual bloggers not being
powerful.
The research, organised around the question why blogs have any (political or societal) influence at all, given their relatively low leadership and lack of central organisation, looks at the skewed distribution of links in the blogosphere, and if this will have consequences for how the blogosphere affects politics (Drezner and Farrell 2008). It was found that there is no central organization to the blogosphere; there is no ideological consensus among its participants (Drezner and Farrell 2008). They argue
9 Zielmann, S. and U. Röttger. Characteristics and developments of political party web campaigns in
Germany, France, the United Kingdom and the United States between 1997 and 2007. In: Baringhorst et al 2009 pp. 69 -‐ 92
that networks in the blogosphere are particularly skewed and this has quite specific implications and influences on the network and network effects.
More specifically, they conclude, many networks are characterised by a distribution in which most nodes have a relatively small number of ties, but a small number of nodes have a disproportionately large number of ties. Therefore, these nodes are not in proportion to the greater total network; Drezner and Farrell explain this as ‘rich’ nodes getting richer. The blogs with the most links get the most attention, thus the ‘rich’ blogs keep getting ‘richer’ in terms of network expansion. These network effects are considered skewed distributions, which ‘have become an important subject of investigation in recent years’ (Drezner and Farrell 2008: 19).
Now, it might look like I am aiming to argue that research to a phenomenon such as the blogosphere has become outdated in a methodological sense, and I would like to take the opportunity here to reflect on this statement. Taking for instance the study by Drezner and Farrell and stretching their approach to fit political communication on Twitter, the argument and research goal of this study might be particularly relevant in the sense of thinking about network topologies on the platform of Twitter. In a sense, the organisational structure of conversation and user engagement on Twitter might be similar to the structure in the ‘old’ blogosphere, while it is being characterised by the same decentralised organisation of conversation, as the sample study by the University of Glasgow concluded.
The research approach as undertaken by Drezner and Farrell is therefore still relevant, as will be discussed in regards with the findings later on. Referring back to Drezner and Farrell’s statement on skewed distributions of nodes and ties in a network on the previous page, is the Twittersphere around the Scottish referendum for example driven by a rich get richer model of network growth, in terms of political
communication, or are there small actors that get (disproportional) amounts of links as well? Put differently, with the blogosphere studies in mind, analysis will be focused at visible network structures in the Twitter sample of the 2014 Scottish referendum Tweets and the implications of the possible skewedness of the data.
With this particular outline of the field of political communication and its characteristics as well as most dominant approaches in mind, I will continue to argue that the field is,
while there are relevant approaches and methodologies, facing the possibility to get stuck in old ways as a research discipline. This argument which will be explored in the next section, dedicated towards justifying and explaining the need for an updated approach towards political communication.
2.2 Contextualisation: towards a multi-‐disciplinary approach
The type of methodologies that are thus generally accepted and have become ‘classic’ in the field of political communication, such as surveys, interviews or direct analysis of content, are generally focused at political advertising and campaigning, such as adverts on TV and in the (mainstream) press. Existing research towards political communication is generally aimed at televised or press political messages (Kaid and Holz-‐Bacha 2006) or more towards studies regarding journalism or media hypes (Louw 2010).
In the case of research to web campaigning, it was found that the emphasis generally lies on the user effects of web campaigning, hence the content layer of web campaigning, rather than the contextual layer of in which it occurs. Although a starting point towards a change of paradigm thus has been slightly established10. Following
Rasmus Nielsen’s argument in the paper Political communication research: new media, new challenges, and new opportunities (2014), I aim to explain why there is a possible need for a change in paradigm.
2.2.1 Demand for shift of paradigm
Keeping the characterisation of the field in mind, it could be argued that research has become relatively homogenous – both methodologically and theoretically, which is currently posing a challenge for future research in the field. Rasmus Nielsen recognises this problem, arguing that the field and its classic methodologies may have remained un-‐ adapted to changes in the media landscape, therefore heading towards, what he calls, an ‘intellectual impasse’ in the field (Nielsen 2014: 6).
However, even though the current methodological challenges, combined with the rise of digital and networked information the field faces sounds like a negative
development for the field, it also offers a lot of opportunity and future possibilities for the field. Overcoming the current challenges are likely going to refresh and update the
field, as long as there will be new methodologies developed that are aimed at sorting out the existing problems.
Put differently, the main troubles of the field consist of homogenisation of research as well as a lack of methodological frameworks to study digital political communication. Nielsen argues that these problems could be overcome by the transformation of the field to become more interdisciplinary, an objective that could be achieved by developing new methodologies in cooperation with research fields in for instance the social sciences, rhetorics, humanities and informatics (Nielsen 2014: 9). He proposes a re-‐ invention of classic methodologies; a more interdisciplinary approach will contribute to achieving this goal, hereby changing the current research paradigm.
Furthermore, Nielsen argues that the field of political communication has a number of approaches or questions that still have not been pursued yet, for instance studying media specificity (how a specific medium and its logic or characteristics contribute to the organisation of communication processes) or research towards how actors make meaning of political communication processes. When operationalising research to political communication in a more media-‐specific manner, the purpose will shift from content analysis towards attention to empirical analysis. As he argues, this is a necessary development, since the classic methodology with the emphasis on
experimental and survey methods ‘construct a research field that is both
methodologically and theoretically relatively homogenous’ (Nielsen 2014: 10). By transforming the classic approaches into a methodology more focused at media itself, researchers will be able to gain a deeper insight in political communication processes, by studying ‘the field’s main publication venues’ (Nielsen 2014: 10).
Overall, it seems reasonable to assume that there is a demand in the field of political communication for a change in methodologies or operationalization of research, and therefore a growing interest in aspects of political communication different from the classic approaches. The main aim in this thesis is therefore to pursue the goal of
developing a new methodological framework, taking the current troubles of the field as presented by Nielsen into account. By moving to a more interdisciplinary approach, which consists in this case of adding theoretical insight of the media studies field with the empirical approaches of the digital methods, it will be possible to gain more insight
in processes of political communication on a media-‐specific level. Hence, this thesis seeks to understand what people do with media instead of what media does with people, by focusing at medium specificity instead of influence and media effects, which will be achieved in a in a critical digital methods methodology featuring analysis of both content and actors as well as the medium (and its critique) itself.
In general, the field of political communication could be considered a quite settled area of research. However, as presented by Nielsen’s argument, the field is currently facing (methodological) troubles. The method as proposed in this paper might therefore be a starting point to assess political communication on the web differently, with its scope more focused on media specificity, taking into account and embrace specific
characteristics of the medium, in this case Twitter, seeing its medium-‐specificity as both a challenge and an opportunity. This medium-‐specific approach will further be
explained and discussed more elaborately in the following section.
2.3 Medium specificity: Twitter and the logic of platforms
An important aspect of this paper is the fact that it seeks to look at political
communication in a medium-‐specific way, hereby focusing at Twitter’s platform logic, its structural organisation and the way this logic influences debate and conversation. In order to achieve a thorough understanding of how the logic of a particular platform organises communication processes, thus how its medium-‐specificity affects this, I will first explain what is meant when referring to platforms throughout this thesis, hence offer a solid definition of what is exactly being studied: a platform.
The definition of a platform has a wide range; one could think of it in terms of an umbrella concept. According to Tarleton Gillespie in Politics of platforms (2010), the term platform has recently emerged to describe online services that are generally based on user-‐generated content (Gillespie 2010). Gillespie argues that the term has a
discursive positioning and is at the same time as a description both specific and vague. Therefore, to narrow the definition of a platform down, Gillespie distinguishes four types of platforms, which align the meaning of a platform together. Firstly, he describes the architectural platform, which is a platform in the most literal sense: a plateau to stand on. Furthermore, a platform could be seen in a computational sense (as
a place to launch software), a figurative sense (a metaphysical platform; such as an opportunity) and a political platform, as a means to express political beliefs and list agenda points. He argues that ‘digital intermediaries depend on all these four’ (Gillespie 2010: 349). In general, platforms could be seen as voice-‐giving surfaces ideas could be launched from.
An important implication of the logic of platforms that should be addressed here is the notion of the ‘walled garden’, a concept coined by Tim Berners-‐Lee in order to make sense of the medium-‐specificity of platforms in general. According to Berners-‐Lee, it is wrong to define platforms as content-‐neutral delivery systems, focused at transferring information without intervention: when content moves from the ‘open web’ to the ‘walled garden’ of a platform, it becomes part of the particular constituted platform-‐ world, with its logic and protocols shaping the social (Berners-‐Lee 2010). This concept is in some sense connected to the idea of platforms actively influencing and shaping power dynamics and social engagement, by limiting it to its set ‘borders’ of what can possibly be done on the platform, as assigned by its particular platform logic. Helmond and Gerlitz identify this phenomenon as ‘the Like economy’, where every action on social is ultimately based on the logical foundation of the platform, the creation of content and its sharing and liking, focused at a data-‐intensive web (Gerlitz and Helmond 2013).
What should thus be kept in mind here when discussing platforms is that they contribute to the organisation of communication processes not just through their medium-‐specificity or their behaviour as a walled garden; platforms are also actively contributing to the process of the ‘engineering’ of the social. According to Helmond and Gerlitz, platforms tend to engineer the social through their specific platform logic, by shaping the (social) interaction that takes place on platforms. Taking a contemporary popular social media platform such as Facebook as an example to explicate this, its platform creates a particular world, a techno-‐social environment in which content is being created, shared and liked, for instance through social buttons (Gerlitz and Helmond 2013). By limiting social engagement to activities that are pre-‐defined, platforms shape power dynamics on the platform. It governs and limits social engagement to the platforms’ specific logic, but in the end it allows the user to post content, share it and like it. One could argue that what we see as ‘the social’ is actually
what the technical infrastructure of a platform allows and prohibits, thus the protocols of a platform. In this sense, medium-‐specificity is actively influencing the actual
communication processes (Gillespie 2010).
2.4 Approaches to Twitter studies: digital methods
When researching Twitter as a platform, one thus should keep the engineered social, the broader social phenomena and societal context in mind. Twitter studies should
therefore be ‘debanalised’: empirical research to the platform and research objectives should be moved away from studying the everyday and the mundane (Rogers 2013: 7). This epistemological change could be carried out by taking the broader societal
concerns regarding Twitter into account, for instance by looking at Twitter as a political platform and as a social phenomenon, with its networks influencing society (Meijas 2010).
‘Twitter I – III’
According to Rogers in the paper Debanalizing Twitter: The Transformation of an Object of Study (2013), an important change in the paradigm surrounding Twitter studies towards debanalisation of the research discipline could be regarded the ‘shift’ of the platform through a number of phases, which he identifies as Twitter I, II and III. By changing its main caption during this process, Twitter has put itself through a transformation, re-‐inventing itself from a medium revolving around the caption question ‘what are you to doing?’ in its early Twitter I days, to the focus on more newsworthy interests in phase Twitter II, moving towards the big data research tool Twitter III (Rogers 2013).
The start of Twitter II could be characterised with the change of the caption phrase to ‘what is happening?’ which caused Twitter’s conversational structure to more newsworthy and relevant subjects; it did not primarily revolve around the mundane, everyday live anymore. This change of the general conversation subject led to an increased level of scholarly attention to the platform; its emphasis on news monitoring prove to provide useful data in scholarly research to elections, disasters and revolution, establishing Twitter as a social research platform (Bruns 2011).
Lastly, Rogers identifies ‘Twitter III’, the phase that is considered to have started after the company entered the stock market. Since entering the stock market, Twitter
data has a particular value and has been fully commodified (Puschmann and Burgess 2014). Twitter III is therefore focused on broader social phenomena with generic data, looking at Twitter as an archived object (Rogers 2013). Twitter III led to a change of paradigm, with a new type of research characterised by an expert use of Twitter, where it is considered a serious research tool and a data market. In general, Twitter could thus be seen as big data machine, giving insight in long-‐term patterns and trends, which might be useful for marketing, prediction and sentiment analysis.
The current environment of the platform with its particular logic and architecture, combined with the data archive of Twitter offers a variety of opportunities and
challenges regarding the current big data debate, a topic that will be further picked up in chapter 2.8. The following chapter, however, will take a number of previous studies into consideration, in order to identify a general understanding of the Twittersphere
specified towards political communication as a research environment, in terms of Twitter as a (news) monitoring tool as well as bigger societal concerns as in the case of big data approaches.
2.5 Approaches to Twitter studies: Similar case studies
Currently, Twitter thus offers a rich environment of (archived) data that can be studied in terms of political communication. As existing studies around Twitter as a political platform show, it can be studied as a medium for political communication in a variety of contexts: it could on the one hand be focused at either the 140-‐character Tweets
themselves, in order to map immediate audience response (Pedersen et al 2014, Larsson and Moe 2012, Bruns and Burgess 2011, Elmer 2013), to predict election results
(Tumasjan et al 2010, Jungherr, Jurgens, & Schoen 2012), to look at the aftermath of a campaign (Mirer and Bode 2010) or, on the other, look at the actual use of Twitter, for instance in 'revolutions' (Rogers 2013) or how Twitter is hands-‐on used by politicians and political campaigns, e.g how a political debate is organised and facilitated by both the public and politicians on Twitter.
With the existing body of research, the road towards studying political communication on Twitter has already been paved by studies specifically dedicated to taking the social network in consideration as a case study. Since the structure of Twitter allows research
to real-‐time politics, whereas 'classic' political communication mediums do not, it was found that existing studies are often paying attention to immediate audience responses, thus research to the public of Twitter. The body of existing research roughly consists of studies to real-‐time events such as conversation during a particular televised debate, or research that aims to analyse Tweets during key campaigning moments such as polls and launches, allowing researchers to gain insight in the 'backchannel' of a political campaign (Kalsnes et al 2014).
Research towards Twitter as a political medium for the public has often -‐similarly to political communication in general-‐ specifically taken American politics into
consideration, for instance by researching the use of Twitter by the US congress
(Golbeck, Grimes and Rogers 2010) or during the presidential elections (Himelboim et al 2012). However, there has also been attention for a variety of Twitter response and conversation around elections and (televised) debates in other countries such as the Nordic countries, Norway and Australia (Larsson and Moe 2010, Larsson and Moe 2012, Kaslnes et al 2014, Bruns and Burgess 2011).
By comparing sets of Tweets sent in Denmark, Sweden and Norway during elections in 2011 and 2012, Larsson and Moe aim to make sense of the conversation on Twitter during a 31-‐day election campaign and seek to identify specific national
differences. By taking particular hashtags into consideration, the researchers try to grasp the entire cycle of a campaign -‐ hereby trying to map audience response around specific events, such as the political discussion in Norway following the 2011 terrorist attack, as well as audience response to, more mundane, key campaigning events (Larsson and Moe 2012: 323).
Bruns and Burgess take a similar approach while studying audience response in
Australia; by looking at a specific hashtag (#ausvote) that is being used throughout the Australian election campaign, the researchers seek to gain insight in audience response during key moments of the Australian elections, and especially aim at looking at how the Australian elections are covered on Twitter.
Kalsnes, Krumsvik and Storsul also proceed to look at to Twitter in the sense of a 'political backchannel', thus the public of Twitter, with research specifically dedicated towards the audience response on Twitter during televised election debates in the 2011
Norwegian election. While engaging with Tweets sent around the televised political debates, the researchers seek to understand what 'backchannel conversation' is featured on Twitter and what sentiment this withholds, thus looking at public conversation and prominent agenda points (Kalsnes et al 2014).
Research that is specifically aimed at mapping the use of Twitter has often been used as a tool to gain further insight in political revolution in the sense of Roger's 'Twitter II'. Rogers mentions in Debanalizing Twitter (2013) a study to the coverage of the
opposition during the Iran elections in 2009, which was nicknamed ‘the Twitter
revolution’11. Despite of the internet censorship in Iran, the researchers aimed to trace
sequences of retweets that had been sent during the elections in Iran, reversing the sorted order of the data, turning Twitter into a ‘narrative’ of retweets about the Iranian revolution (Rogers 2013).
As much attention has been paid to the use of Twitter as a political tool for just the public and a means to engage the public, allowing researchers to gain insight in public sentiment and opinion, in a more stable political environment (as opposed to what was the case in Roger's Iranian study), there has not been much attention towards the use of Twitter as a campaigning tool for both politicians and users. Research towards Twitter in the context of political communication has thus been mostly focused at the public of Twitter and their use of it as a political medium; at the Tweets and the content that they feature themselves, as well as the occurrence of Tweets during specific moments or the lack of them. There has not been much attention paid to the hands-‐on use of Twitter and the attributes of the platform by both the public and politicians – the use of specific hashtags, mentions or retweets -‐ in a campaign and how Twitter can be used by both politicians and public to organise and facilitate a political campaign, yet. Especially in the sense of the Scottish referendum, specific research to this topic has not been (fully) operationalised yet or is still in the draft or reviewing phase of publishing.
Published research that is only focused at the topic of the Scottish independence
referendum is thus (yet) of a limited amount; A relevant study that was found also aims to map the responses on Twitter to a televised debate about the referendum (Pedersen
et al 2014). The researchers seek to look at peaks and troughs in the amount of Tweets sent during a televised debate and what specific language and hashtags are being used in the Twitter debate, to make sense of the 'backchannel' of a political campaign (Pedersen et al 2014). Again, by looking at immediate audience response, this study can be
considered a study towards the public use of Twitter, since it seeks to make sense of the public opinion on Twitter during a televised debate.
The research I aim to pursue here is focused at the actual use of Twitter by both
politicians and users, in order to organise a political campaign as well as a debate. Based on the conclusions of the research undertaken by the University of Glasgow in 2013, it seeks to understand how the campaigning debate around the Scottish referendum is being organised on Twitter. By taking the Glaswegian claims into account, this thesis will go beyond the 'Twitter as backchannel' paradigm that is characterised by studying public opinion. It seeks to take the organisation of the campaign and the facilitation of it by both politicians and sympathising users of each side of the campaign into
consideration, by looking at network behaviour of content and users, as well as the actual content itself. Therefore, this thesis is focusing at both the public use of Twitter as well as the political use of it.
Since my research will thus focus on networked conversation and the ‘stretch’ of the official campaigning debate over the more general conversation around the Scottish referendum on Twitter, thus the networked structure of Tweets and its implications, the next chapter will introduce some theoretical approaches towards the notion of a
network and the theory around how networks and their topologies are being organised.
2.6 Networks and topological theory
It might be interesting to look at the networked structure of Tweets; due to the decentralised structure of the platform through its platform politics, there is a
possiblility of a skewed distribution of connections and nodes within the conversation network. The study to political communication in the blogosphere by Drezner and Farrell, as was discussed earlier, claims that the networked structure of the blogosphere is skewed in a particular way, which has quite specific effects on the organisational structure and thus the blogosphere network and network effects (Drezner and Farrell 2008).