SM&S 2024

The Twitter That Was: Reflections on Ten Years of #auspol

Axel Bruns and Anand Badola

• 18 July 2024 – Paper presented at the Social Media & Society 2024 conference, London

Presentation Slides

Abstract

Introduction/Background

In the wake of the takeover of Twitter (now ‘X’) by Elon Musk and his financial backers in October 2022, Twitter as a platform has declined in vibrancy and relevance in general, and Twitter research in particular has become virtually impossible as the Academic API was closed down and replaced by a paid access point that has proven unaffordable for ordinary researchers. Alongside a broader trend in the social media landscape towards less observable platforms, this is concerning for public-interest research that seeks to investigate processes of public debate and public opinion formation and to analyse the involvement of legitimate political actors as well as problematic influence operations in such processes. However, this distinct moment of disruption also serves as a useful prompt for reflection about the role that Twitter has played as a platform for public political debate during its heyday over the past decade or more.

Drawing on a unique longitudinal dataset of a consistently highly active and vibrant online community on Twitter, this paper examines the evolution of political debate on the platform. Covering the period from December 2013 to July 2023, it focusses on the long-standing political hashtag #auspol (cf. Sauter & Bruns, 2015), used for tweets relating to Australian politics at the federal level, which has been the single most persistent and active hashtag in the Australian Twittersphere (Bruns & Moon 2019) since Twitter was widely adopted as a platform for public communication in the country. Hosting a central core of consistently engaged users – whom we may understand as ‘political junkies’ (Coleman, 2003), usually without formal political roles – surrounded by a broader group of less engaged participants, including politicians, journalists, activists, and other political stakeholders, and widely visible across the Australian and international Twittersphere as a result of retweeting and the trending topics algorithm, #auspol can be understood as a genuine community of participants in a cultural studies sense (cf. Baym, 1998); however, such comparatively stable community structures were regularly disrupted by a new and often temporary influx of additional participants during political scandals, crises, elections, and other periods of heightened political attention and debate. Especially during such periods, #auspol also became a ‘participatory object’ (Marres, 2012), well beyond Twitter itself, as participants in Australian political debates (including in political statements and journalistic commentary) began to refer to it as shorthand for a perceived vox populi – or at least a vox Twitteratorum, as Sauter & Bruns (2015) put it.

Objective(s)/Research Question(s)

Understanding #auspol in the first place as a community of ‘political junkies’ discussing the events of the day in Australian federal politics, this paper centrally examines the structure and dynamics of that community. Informed by cultural studies understandings of online community structures, it distinguishes between a central core of highly active and persistently committed community members as well as less engaged occasional and transient participants, and explores their contribution to the total debate over time. Over the ten-year timeframe of the study, it also investigates fluctuations both in the overall make-up of the core group of community members, and in their positioning towards each other and towards broader Australian political camps. These analyses serve both to support the central claim that #auspol is indeed a community in the narrower sense of the term (and not merely a cluster of frequently interacting nodes in a network), and to demonstrate that the structure of that community is itself dynamic over the extended timeframe of the present study.

Further, the gradual changes in the structure of the community provide additional insight into the presence of polarising or depolarising tendencies within Australian political debate at various points in the timeframe (as driven, for instance, by debates and controversies on specific major issues, or by the rhetoric of particular political actors), as well as into the impact of various platform interventions (such as user suspensions or readmissions). Additionally, the analysis also pays special attention to the additional accounts engaging in #auspol during periods of heightened attention (including elections and political crises), and the extent to which such accounts are recruited into longer-term participation as a result of these experiences.

Additionally, this study also examines the shifting themes and topics of Australian political debate over the course of these ten years. Analysing patterns in the tweet texts, the use of secondary hashtags, and the inclusion of external links to news outlets and other sources, it explores the changing emphases in Australian political debate, showing the relative importance of recurring debates on migration, climate change, and taxation policy as well as the impact of shorter-term and unexpected issues such as the COVID-19 pandemic. Examining these distinct debates for themselves, the study investigates whether the agonistic and antagonistic alliances between participants remain stable or change from issue to issue, and thereby explores whether the political faultlines between them are based predominantly in specific issue-based or general ideological polarisation (cf. Esau et al., 2023).

Method

The dataset underlying this analysis is the result of a long-term effort to collect #auspol tweets. Using a variety of tools connecting to the various versions of the Twitter API available at the time, this has compiled a virtually uninterrupted stream of tweets from 16 December 2013 to 29 July 2023. Tweets were collected contemporaneously throughout this period, rather than through the archival Twitter Academic API at a much later date, and the dataset therefore contains all tweets available at the time of collection, without the degradation in fidelity (through tweet or account deletions and other factors) that is common in archival datasets collected well after the fact. The dataset is shaped by the changing API frameworks employed by Twitter over the past ten years, and has been constructed and standardised from several partial datasets in distinct collection formats; however, this has not affected the core data fields (tweet ID, text, sender account information, timestamp) that the present analysis centrally draws on.

The study processes this dataset using a number of analytical approaches. In addition to long-established activity measures for Twitter data (cf. Bruns & Stieglitz, 2013), it employs network analysis techniques to determine and visualise community structures and interaction patterns, as well as topic modelling and topic similarity approaches to identify dominant themes and correlate them with network patterns. These analytical techniques are applied to the full dataset, but the analysis also introduces the dimension of time in order to trace their evolution over the course of the ten years of activity covered by the dataset.

Preliminary Results

The full dataset contains some 57 million tweets from over 1.2 million accounts; this represents an average of over 16,000 tweets per day for the more than 3,500 days covered by the timeframe, and an average of some 46 tweets per account. As is common in social media data, however, such activity is very unevenly distributed: while the median number of tweets per account over this timeframe is 1, a small group of accounts in the dataset contribute the vast majority of tweets. Indeed, the top 1% of most active accounts (12,452 accounts in total) produced nearly 45 million of all tweets (78%); the next most active 9% of accounts contributed another 10 million tweets, so that the top 10% of accounts together posted nearly 96% of all tweets. This demonstrates that there is a small and very persistent core of highly active accounts in the #auspol community; an average of some 2,600 unique accounts within this core community actively posted to #auspol on each day covered by the 10-year timeframe; 5,200 actively posted each week.

Fig. 1: @mention network for the entire time period (size: weighted indegree, colour: outdegree)

Such extensive commitment may appear to indicate a shared sense of purpose, but this is true only as far as overall commitment to debate is concerned; indicators of political division emerge even from the preliminary analysis. A network visualisation of @mentions between accounts for the entire ten-year period (fig. 1) shows considerable engagement between participants, but already also hints at the existence of distinct subgroupings within the community (a core group at the centre, and several smaller clusters to the top and right); while a visualisation of retweets between accounts produces a significantly stronger subdivision between a smaller but tightly packed cluster on the left and a much larger and more loosely organised cluster on the right (fig. 2). Retweets (which often represent a form of endorsement or amplification) are more likely to show issue-based or ideological alignment, while @mentions may also address antagonists in order to debate or criticise them; these divergences in the overall network structure therefore appear to point to the presence of long-standing cleavages in the #auspol participant base. Smaller-scale distinctions are likely to emerge from further analyses focussing on specific timeframes within the overall ten-year period.

Fig. 2: retweet network for the entire time period (size: weighted indegree, colour: outdegree)

Future Work

Like major global hashtags such as #metoo or #BlackLivesMatter, #auspol has arguably managed to transcend Twitter itself as it found recognition well beyond the platform; unlike these other hashtags, however, it sparked no political movement and remained grounded in Twitter as a space for public debate. This is likely due to the make-up of its userbase: its primary function remained as a mechanism for bringing together a diverse community of participants in continuous – if not always constructive or respectful – discussion about the political topics of the day, rather than for organising collective (or, indeed, connective; cf. Bennett & Segerberg, 2012) action on distinct issues or injusticies. In this, it is unusual if not entirely unique (some other national politics hashtags, like the Canadian #cdnpoli, have served similar purposes), and its longevity and persistently high level of activity mark it out as deserving further attention.

This abstract permits space only for very basic preliminary analyses, yet they already point to the considerable insight that further analysis can extract from this unique longitudinal dataset. The full paper will present a substantially more detailed study of the community structure and activity dynamics of #auspol between 2013 and 2023, providing a thorough review of this long-standing phenomenon and its co-evolution with overall political debate in the Australian public sphere. In doing so, it will reflect on the role that #auspol has played in Australian politics, and, by extension, on the role that Twitter played as a public platform for public debate overall – a role that has declined precipitously following the Musk-led acquisition and gutting of the company, and whose vacancy has significant consequences around the world.

References:
Baym, N. K. (1998). The Emergence of On-Line Community. In Cybersociety 2.0: Revisiting Computer-Mediated Communication and Community (pp. 35–68). SAGE Publications, Inc. https://doi.org/10.4135/9781452243689

Bennett, W. L., & Segerberg, A. (2012). The Logic of Connective Action. Information, Communication & Society, 15(5), 739–768. https://doi.org/10.1080/1369118X.2012.670661

Bruns, A., & Moon, B. (2019). One Day in the Life of a National Twittersphere. Nordicom Review, 40(s1), 11–30. https://doi.org/10.2478/nor-2019-0011

Bruns, A., & Stieglitz, S. (2013). Towards More Systematic Twitter Analysis: Metrics for Tweeting Activities. International Journal of Social Research Methodology, 16(2), 91–108. https://doi.org/10.1080/13645579.2012.756095

Coleman, S. (2003). A Tale of Two Houses: The House of Commons, the _Big Brother_ House and the People at Home. Parliamentary Affairs, 56, 733–758. https://doi.org/10.1093/pa/gsg113

Esau, K., Choucair, T., Vilkins, S., Svegaard, S., Bruns, A., & Lubicz, C. (2023, May 30). Destructive Political Polarization in the Context of Digital Communication – A Critical Literature Review and Conceptual Framework. International Communication Association, Toronto. https://eprints.qut.edu.au/238775/

Marres, N. (2012). Material Participation: Technology, the Environment and Everyday Publics. Houndmills: Palgrave Macmillan.

Sauter, T., & Bruns, A. (2015). #auspol: The Hashtag as Community, Event, and Material Object for Engaging with Australian Politics. In N. Rambukkana (Ed.), Hashtag Publics: The Power and Politics of Discursive Networks (pp. 47–60). Peter Lang.