RIP #almedalen: The rise and fall of the hashtag of a Swedish democracy festival

Abstract

Democracy festivals are events that aim to strengthen democracy by engaging citizens, politicians and organisations in dialogue and provide a more equal access to the agenda. In recent years, social media have become important arenas for agenda building, supposedly equalising the access to such processes. This article uncovers patterns of activity and visibility in agenda building through an actor-centric study of the official X/Twitter hashtag of a Swedish democracy festival. We analyse all tweets containing the hashtag #almedalen for four election years and find that levels of activity in the hashtag are connected to the visibility of elite actors. We see a cycle of development concerning visibility in the hashtag over time with early adopters dominating the beginning of the period, establishment actors in the middle and right-wing populist actors at the end. There is little evidence of the hashtag as an equalising factor in agenda building. We discuss these findings in relation to the position of X/Twitter as an arena for opinion formation.

Democracy festivals, democratic agendas and dissonance

Many remedies have been proposed for the ailments of representative democracy, including revitalising the role of citizens in politics through democratic deliberation in the public sphere (cf. Dryzek et al., 2019; Habermas, 1991). In the 21st century, such a task has often been envisioned as being aided by digital and social media such as Twitter ¹ (cf. Loader and Mercea, 2011). However, digital communication platforms seem to produce what Pfetsch (2018) has labelled ‘dissonant public spheres’, characterised by contestation, polarisation, hate-speech, disinformation, noise and confusion, as the traditional gatekeeping role of legacy media declines in significance and non-institutional actors enter the hybrid media system (Chadwick, 2017).

By focusing on actors in social media in relation to a specific event over time, we intend to uncover patterns in activity and visibility in the agenda-building process, defined as ‘the process by which demands of various groups in the population are translated into items vying for the serious attention of public officials’ (Cobb and Elder, 1971: 126). Our empirical setting is the official hashtag of a democracy festival, since this presents an arena which is important for central actors while still broad enough to capture a great number of different issues. Moreover, such festivals are inclusive enough to capture the interest of non-elite actors. We assess the visibility and centrality of specific actors in the hashtag over time and explore emerging patterns of user popularity. We find that the attractiveness of the platform and the event corresponds to changes in the attractiveness of the hashtag, and that resource-rich establishment actors will dominate more than non-establishment actors when the hashtag is more attractive. In addition, our results show traces of the successful incursion of far-right policy actors into the Swedish public sphere in the late 2010s – suggesting tendencies of development towards a more dissonant public sphere as defined above.

To the best of our knowledge, this study is the first to provide an exploratory study of actors in the agenda-building process on social media. Moreover, it is also unique in that it studies the use patterns of a single hashtag over an extended period.

Democracy festivals are events held with the stated intent to ‘revitalise democracy by connecting the political system with citizens and their organisations, enhancing dialogue and participation, and ensuring access to core political discussions and actors across resource differences’ (Wollebæk and Raknes, 2022: 334). Empirical studies have, however, shown that democracy festivals tend to reproduce resource differences, as shown by analyses of such events in Norway and Denmark (Sørensen and Lund, 2019; Wollebæk and Raknes, 2022). Nevertheless, these studies surveyed experienced networking and legacy media coverage benefits of formal organisations rather than social media activity, where resources might play a smaller role. Indeed, it has been suggested that social media can be a way for comparably resource-poor policy actors to gain influence (Ihlen et al., 2022).

The first democracy festival was Almedalsveckan, (‘Almedalen week’, henceforth Almedalen; Östberg, 2013: 14). Almedalen has inspired counterparts in the other Nordic countries: Suomiareena (Finland 2006), Folkemødet (Denmark 2011), Arendalsuka (Norway 2012) and Fundur Fólksins (Iceland 2015), and later in Turkey, the United Kingdom and Hungary. Considering the high-level elite attendance and the big media attention that Nordic democracy festivals attract, there has been surprisingly little research devoted to them (Wollebæk and Raknes, 2022). A democracy festival devised to create equal opportunities for influencing the agenda, aided by the equalising factors of social media, and set in Sweden, one of the most egalitarian countries in the world, could be a most-likely case for socially inclusive agenda building (Boräng and Naurin, 2022). Considering the #almedalen hashtag as an integral part of the democracy festival, we expect to find a broad mix of elite and non-elite interests visible in the hashtag if there is socially inclusive agenda building, and a dominance of established, resource-rich policy actors if the opposite is the case. Before moving on to assessing our hashtag data, we describe the attention enjoyed by Almedalen in Swedish society during recent years. Figures 1 and 2 provide such overviews.

Figure 1.

Number of articles in Swedish print media per year containing the word ‘Almedalsveckan’, 1990–2022.

Figure 2.

Number of official Almedalen events, 2001–2022.

As shown in Figure 1, the 2000s saw a large increase in media attention for Almedalen, culminating in the election years of the 2010s (2010, 2014 and 2018). The intensive news reporting and public debates surrounding Almedalen were shown to have a crucial effect on political agenda-setting processes (Östberg, 2013; Wendt, 2012). In the 2018 election year, 4311 official events (speeches, seminars, debates etc.) were arranged by 1929 separate actors and attended by an estimated number of 45,000 unique visitors (Almedalsveckan, 2018). This would prove to be the peak year (see Figure 2). There were considerably fewer organisations and events in the following post-election year, and in 2020, the event was cancelled due to the coronavirus pandemic, which also had a dampening effect in 2021. In the election year of 2022, there were 2100 events arranged by 1100 separate actors attended by 19,000 unique visitors, that is, only half of the 2018 numbers (Almedalsveckan, 2022).

Thus, while still an important event in Swedish politics and civil society, the data presented in Figures 1 and 2 suggest that the heyday of Almedalen has come and gone. We will not dwell on the reasons for its apparent demise. Instead, we utilise the ‘rise and fall’ of the hashtag to study the centrality of policy actors. Influencing the political agenda during an election year is essential, both for candidates and parties, as well as for interest groups trying to further their causes. In this way, studying the hashtag during election years will allow us to study it at its height of policy impact, and, conversely, at the height of interest from actors engaging with it (cf. Seethaler and Melischek, 2019). Thus, we propose the following overarching research question:

• RQ: What characterises the development of the #almedalen Twitter hashtag from 2010 to 2022?

We choose to approach this from an actor-centric perspective. This will help us identify networks of actors and which actors are central to these networks. The most central actors in the hashtag are not necessarily the most influential ones in terms of agenda building, but we propose that the centrality of actors is indicative of which actors are most successful in dominating discourse in the hashtag, which is one aspect of agenda building. We thus posit the following subquestions.

• SQ1: Who are the dominant actors in the hashtag in terms of visibility for each year?

• SQ2: How are the dominant actors in the hashtag connected to each other in terms of interaction?

Agenda building through Twitter hashtags

Agenda building and social media

Agenda building ‘explores the sources that make up news content and influence the mass media agenda’ (Tedesco, 2011: 78). Journalists strive for independent sources but often rely on policy actors’ PR, which strategically shapes information. They use so-called information subsidies (Gandy, 1982) – all types of promotional materials such as ads, posters and press releases – and try to influence journalists by ‘reducing the journalists’ costs of gathering information’. In other words, policy actors ‘subsidise’ information for journalists (Berkowitz and Adams, 1990: 723).

The processes of agenda-setting in legacy media can be roughly characterised as the outcome of mass media logics (Altheide and Snow, 1979), which explain how media outlets shape their depiction of events and issues out of the need to attract viewers and readers, and the mediatisation of politics (Mazzoleni and Schulz, 1999). This illustrates how political actors react to this logic by shaping their communication to fit within a media-centric system.

Social media, by contrast, functions in partly different ways, as expressed by Van Dijck and Poell’s (2013) notion of social media logic. Here, visibility is driven by the algorithmic reinforcement of popular content (McCombs et al., 2014). Social media also expands the inventory of information subsidies with different types of social media posts (tweets, TikTok videos, Instagram posts etc.). Following Parmelee (2014), we here consider tweets to be a form of information subsidies. Crucially for this study, social media introduces a more complex network of policy actors in the agenda-building process. The concept of agenda-setting has previously been expanded to include not only the transfer of issue salience from the media to the public, but also from the media to political actors (Van Aelst et al., 2014). Furthermore, it has been observed that whereas public conversation on social media often revolves around legacy media stories, likewise legacy media can cover an issue that is especially salient or dominant in social media. Drawing on Chadwick’s (2017) concept of the political information cycle, and the complex network of actors involved therein, we argue that agenda-building processes taking place in multi-platform arenas can potentially transfer issue salience from policy actors to the public; from the media to the public; from the media to policy actors; from the public to the media; and from the public to policy actors. Thus, social media like Twitter makes it possible for actors to at least partly bypass the gatekeeping function of legacy media (Parmelee, 2014; Seethaler and Melischek, 2019).

Given the complex media landscape with diverse actors and competing agendas, it is surprising that agenda-building research has not focused more on which actors shape and dominate the process. To fully understand how an evolving media system affects agenda-building processes, it is necessary to study the visibility and interaction of individual policy actors. In a traditional media system, agenda-building success depends on legacy media’s interpretative dominance, ensuring an actor’s information prevails over competitors’. For the purposes of the present study, we posit that one of the core functions of the #almedalen hashtag is indeed agenda building, and that high visibility of an actor within that hashtag can be understood as a form of interpretative dominance (cf. Kanervo and Kanervo, 1989).

Twitter as a space for agenda building

Despite its small user base, dominated by journalists, political actors, activists and opinion leaders (Hedman, 2019), Twitter played a key role in Sweden’s public sphere during the 2010s and early 2020s, serving both as a social platform and an ‘ambient information stream’ (Bruns and Burgess, 2012). From around 2% of Swedish Internet users being daily Twitter users in 2010–2011, the level has been remarkably stable at 6–7% since 2013 (Andersson et al., 2022).

Swedish politicians adopted Twitter quickly. By 2013, more than half of Swedish MPs had Twitter accounts (Larsson and Kalsnes, 2014). Politicians use Twitter and other social media to communicate with voters and journalists, raise money, increase their reach, test new communication tactics and keep track of public opinion (Severin-Nielsen, 2023). Enli and Simonsen (2018) summarise politicians’ uses of social media as encompassing political marketing, agenda-setting, and internal communication. Indeed, politicians have been shown to use hashtags in strategic ways to increase their visibility (Darius, 2022).

Relatedly, Swedish journalists embraced Twitter from early on (Djerf-Pierre et al., 2016). Already by 2011, more than half of Swedish journalists were reported to be at least passive users (Djerf-Pierre, 2012). Journalists use Twitter and other social media platforms for various purposes, such as following discussions, trendspotting, research and getting feedback (Enli and Simonsen, 2018; Hedman, 2019). Hashtags help journalists to find information while doing research but can also be used to increase their own visibility (cf. Enli and Simonsen, 2018).

Beyond politicians and journalists, Twitter is employed by a diverse set of policy actors seeking to influence the public, political and media agendas: lobbyists, interest groups, social movements and individual activists. Such advocacy groups tend to use Twitter to gain attention from journalists and ‘to try to enrol politicians as proxy advocates for one’s positions’ (Johansson et al., 2019: 1536), whereas other social media platforms are used to reach a wider audience.

In agenda building, the relative influence of actors will be determined by a combination of their formal positions and their skills in social media tactics (cf. Enli and Simonsen, 2018). Twitter was originally considered to be a promising space for interaction between politicians and ordinary citizens, an inclusive public sphere that would make elites more accessible (Rogstad, 2016). However, empirical research suggested the contrary: small elite groups of politicians, journalists and opinion leaders interacting with each other in small clusters (Darius, 2022; Enli and Simonsen, 2018; Larsson and Ihlen, 2015), in front of a larger audience.

As noted above, studies indicate that democracy festivals seem to favour resource-rich interests (Wollebæk and Raknes, 2022). We thus expect that increased use of the Twitter hashtag will lead to the dominance of establishment actors, whereas lower use of the Twitter hashtag will provide more opportunities for non-establishment actors. Twitter as a space for agenda building has also been used by far-right political actors, often successfully using popular hashtags to influence public discourse (Åkerlund, 2020; Larsson, 2014). We thus foresee that our results will demarcate a potential struggle between establishment policy actors and non-establishment actors in the hashtag, thereby bringing about a more dissonant and segregated pattern (Pfetsch, 2018).

Methods

Data collection

Data was collected by means of the now-defunct Twitter API for Academic Research, allowing access to the full history of Twitter (2021). The academictwitteR package (Barrie and Ho, 2021) was used to archive all tweets that included the #almedalen hashtag and that had been posted during the Almedalen events during four Swedish national election years – 2010, 2014, 2018 and 2022. Focusing on hashtags was deemed suitable as this mode of communication was foreseeable from year to year (e.g., Bruns and Stieglitz, 2012). While we can be certain that content related to Almedalen without the hashtag was tweeted during our studied time periods (e.g., Lorentzen and Nolin, 2017), previous research suggests #almedalen to be an established hashtag (e.g., Larsson and Christensen, 2016). Our focus on election years (as noted above) is motivated by the heightened levels of interest in political matters during such time periods (Larsson, 2017, 2019). Details regarding the dates covered and the number of tweets collected can be found in Figure 3.

Figure 3.

Number of tweets per year (upper section), per year and day (middle section) and types of tweet per year and day (lower section).

Data analysis

Social media platforms like Twitter change over time (e.g. Helmond et al., 2019). Certain types of Twitter usage have, however, remained more or less constant throughout our studied period. Specifically, we can differentiate between four types of tweets: original tweets, mentions, replies and retweets. For original tweets (terminology suggested by Bruns and Moon, 2018), this type of tweet can perhaps be primarily defined by a scarcity of interactivity markers (featuring various uses of the @ character as discussed below). Original tweets can be thought of as tweets posted to the platform with no specific recipient in mind. Mentions and replies are quite similar to each other. Both types make use of the @ character in combination with a username to indicate an attempt at interaction. Certainly, the @ character has been referred to as representing the ‘primary interactive affordance’ (Hemsley et al., 2018: 1) of the platform at hand. Our data collection approach allows for the distinction of replies from other types of tweets as a message sent in reply to another tweet. Tweets not classified by the API as replies, but that include one or more @ characters in relation to one or more usernames are considered mentions. Thus, while replies are frequently seen as part and parcel of conversations on Twitter, mentions are often utilised to ‘tag’ other users – a practice which in turn can lead to discussions via replies. Our main focus for the study at hand, however, is placed on retweets. Allowing users to redistribute tweets originally sent by others, retweeting can be likened to the share functionality featured on Facebook (Larsson, 2015). For the user being retweeted, such attention can be quite important as redistribution is essential to reach virality (Klinger and Svensson, 2018). For the user sending the retweet, the act of redistributing can be construed as a linkage between one’s own profile and that of the retweeted user (Venturini et al., 2018). While we cannot know for certain the inner workings and reasonings behind each sent retweet, redistributing content in this way helps boost the content and the user behind the content, which allows us to study the visibility and centrality of elite and non-elite actors in the agenda-building processes.

We employ the network mapping tool Gephi to map out retweeting patterns between users of #almedalen for each studied year. The resulting network maps can be found in Figures 4 to 7. Each map features a series of nodes, each representing a specific user. Node size is dependent on the indegree measurement, which in this case refers to the extent to which a specific user has been retweeted by other users. A larger node indicates larger amounts of retweets enjoyed by the specific user. Node placement is based on application of the ForceAtlas2 algorithm, which structures networks so that highly connected nodes are placed closer to each other, while less connected nodes are pushed further apart (Jacomy et al., 2014). After filtering out nodes with comparably few retweets, colour was assigned to the nodes in order to further identify clusters of users whose retweeting patterns were similar. Specifically, node colour was determined by the modularity class algorithm offered by Gephi, which colours nodes that have more connections similarly to allow for the identification of potential clusters or subnetworks of users (e.g., Bruns et al., 2017). Once constructed, the network maps featured in Figures 4 to 7 allowed us to identify users whose nodes were comparably larger than others – indicating that the corresponding users had received more attention in terms of gaining retweets. Once identified, the tweets sent by these large-node users – and the user profiles themselves – were sought out and subjected to qualitative close readings focusing on the connection between them and the resulting characteristics of the cluster formed as a result of their activities (drawing on, for example, Bruns and Moon, 2018).

Figure 4.

Highly retweeted users in the #almedalen hashtag during the 2010 event.

Figure 5.

Highly retweeted users in the #almedalen hashtag during the 2014 event.

Figure 6.

Highly retweeted users in the #almedalen hashtag during the 2018 event.

Figure 7.

Highly retweeted users in the #almedalen hashtag during the 2022 event.

Results

Figure 3 provides some descriptive statistics about the volume and type of activities undertaken in the #almedalen hashtag.

The upper section of Figure 1 indicates that in terms of pure numbers of tweets, our first studied period (in 2010, with 24,846 tweets) saw comparably modest use of the hashtag – with even less frequent use taking place during the final period studied, 2022 (7989 tweets). The bulk of activity is instead focused on the 2014 (124,203 tweets) and 2018 (80,513 tweets) election years. This could indicate a change in the attractiveness of the Twitter hashtag as an agenda-building platform, possibly exacerbated by a change in the attractiveness of the Almedalen event itself. In the following, we shall see how the changes in activity in the hashtag accompany the retweet networks of policy actors.

Much like activity wanes over our four studied years, so the level of activity undertaken in #almedalen varies over each period. The middle section of Figure 3 provides insights into these tendencies. As the 2014 and 2018 events saw most use of the hashtag, we consequently see the curves representing these years stretch out further over the dates for the events. For 2010 and in particular for 2022, we see activity peaking during the beginning of each respective period, after which the corresponding lines indicate a more rapid decrease of activity than is visible in the trajectory for the lines representing the 2014 and 2018 periods.

Finally, the lower section of Figure 3 suggests that not only does the level of activity change over the studied time periods – the type of activity undertaken in #almedalen changes as well. While the lines featured in the lower section for the 2010 event indicate a dominance of original tweets (61%, n = 15,308 throughout the period), later years instead see hashtag use shift towards retweets. Indeed, with retweets constituting 53% of tweets in 2014 (n = 66,018), 64% of tweets in 2018 (n = 51,692) and 65% of tweets in 2022 (n = 5126), the lower section of Figure 1 suggests a clear development towards redistributive activity rather than dialogue and discussion.

Next, we move on to assess the users and groups of users that gained comparably higher amounts of attention in the #almedalen hashtag during our four studied periods and which can be used as an indication of which actors in the agenda-building process achieve high visibility and centrality. Figure 4 details the retweeting activity undertaken in 2010.

Identifying separate clusters of users based on their retweet patterns proved difficult for the 2010 data. While modularity class was calculated at different resolutions, these calculations typically led to the identification of 50 or more clusters. Thus, we consider the network for the 2010 Almedalen week to be ‘structurally ambiguous’ (Venturini et al., 2021, p. 12), thereby not lending itself very well to cluster identification. Keeping the rather small number of users in 2010 in mind (cf. Figure 3), we suggest viewing the structure visible in Figure 1 as one cluster of users, more interrelated than the users found in different clusters for later years (as depicted in Figures 5 to 7). While no clear clusters could be identified by means of modularity class calculation, we can nevertheless identify some clustering tendencies in Figure 1. First, the upper-left-hand section of the figure sees a large node representing the user mpawlo. Describing himself as an entrepreneur on his Twitter profile page, mpawlo gained attention in #almedalen in 2010 by urging political bloggers to let their voices be heard regarding laws to be passed on intellectual property rights. Perhaps unsurprisingly, the retweets enjoyed by this user appear to primarily emanate from the upper half of Figure 3, featuring communications professionals (britstakston, jocke, nikkelin, deeped, mickep2, westander, doktorspinn) and users related to a broader sphere of (primarily centre-right) politics (federley, makthavare, hannawagenius). Much like the case with mpawlo, the lower section of Figure 1 features a left-hand side node primarily related to centre-left and left-wing political actors (ulfbjereld, annaardin, aliesbati, feministerna, martaaxner, miljopartiet, vansterpartiet) and journalists and media organisations (matsknutson, svtvalet2010, emanuelkarlsten). Specifically, karinadelskold who presents herself as a ‘journalist, comedian and author’ on her Twitter profile (Adelsköld, 2023) appears to have gained considerable attention based on her tweets surrounding an event during the 2010 week where the then-leader of the Feminist Initiative party literally burned 100,000 SEK to illustrate gendered salary differences (SVT, 2010).

Finally, the middle section of Figure 1 features a small node with a series of outgoing connections. The corresponding user is annoula64 – a Pirate Party supporter who has exhibited similar tendencies to frequently engage in retweeting other users in relation to Swedish elections (Larsson et al., 2016; Larsson and Moe, 2012, 2016) and current affairs talk shows (Larsson, 2013).

For the 2014 Almedalen, tweets carrying the studied hashtag suggested three distinct clusters of users. The middle section of Figure 5 is in many ways reminiscent of the tendencies uncovered for the 2010 event. For instance, we again find examples of accounts suggesting primarily left-wing or liberal political affiliations (spirande, ulfbjereld, socialdemokrat, feministerna) as well as journalists and media organisations (emanuelkarlsten, dagenssamhalle). Moreover, while individual communications professionals (britstakston, westander, hampusbrynolf) can be found here as well as in 2010, the 2014 event sees a clearer influence of companies offering PR and communication services (notifiedswe, heylissly). Similarly, in what could be labelled as a professionalisation of #almedalen activities, the middle section suggests the increased influence of organisations such as trade unions and professional interest groups (losverige, lararforbundet, facketkommunal, vardforbundet, skommunikatorer), employer organisations (almegaab, friskola), as well as non-profits (rodakorset, barnibehov, rfsu, unicefsverige). Another indication is the presence of the Almedalsguiden (literally ‘Almedalen guide’, author’s translation) account. The profile describes itself as ‘the provider of the official Almedalen program since 2013’ – but we also see (at the time of writing) that the latest activity from the account appears to be a retweet posted in 2019 (Almedalsguiden, 2024).

One key user in the middle section of Figure 5 exhibits clear connections to the right-hand side of the same figure. Specifically, looking closer at the retweets emanating from ozznujen, we can see the comedian Özz Nûjen, who gave a performance in Almedalen of a play in which he acted as prime minister, at the same time as Sweden Democrats leader Jimmie Åkesson gave his speech from the main stage (Sandberg, 2014), succeeding in gaining retweets by users across the aforementioned blue, middle cluster, as well as by users in the rightmost orange-coloured cluster. While this latter section again features journalists and media organisations (niklassvensson, u_kristofferson, expressen) the most popular accounts here belong to right-wing politicians (oliverrosengren, denniswedin, carlbildt) and parties (nya_moderaterna, alliansswe, kdriks). niklassvensson was the most prolific political journalist in Sweden at the time, known for using Twitter extensively for self-promotion and building relationships, becoming a ‘reinvented journalist’ on Twitter (Hedman, 2019).

Much like for 2010, the 2014 event also spawned one user who appeared to be primarily active by means of retweeting others. In 2014, this role appears to have been inherited by beckmansasikter, an account operated by conservative MP Lars Beckman.

For 2014, finally, the lower left-hand side of Figure 5 features a cluster identified by green colour. Users here appear to have climate-related issues in common — the graph allows for identification of electricity and energy companies (eon_nordic, fortumsverige, haga_initiative) as well as of users associated with Greenpeace (rolflindahl, isadora_wronski). This is a further indication of how professional, more institutionalised policy actors move into the arena as it becomes more attractive for agenda building.

The year 2018 saw a lower number of tweets in the hashtag compared to the peak year of 2014, as well as a higher share of retweets (cf. Figure 3). Compared to 2014, there is also a decreased importance of communication professionals and an increasing visibility of right-wing accounts. Starting with the green cluster visible to the left-hand side of Figure 6, one user emerges as especially important for those engaging with #almedalen in 2018: truthandfiction, operated by journalist Annika Hernroth-Rothstein who tweeted about the KRIS organisation (ex-convicts working for meaningful roles in society) blocking a Nazi march. Indeed, the KRIS Twitter account – rikskris – is clearly visible in the green cluster, suggesting a considerable amount of online activity emanating from this dramatic event. Other users with large nodes also touch upon this event. For instance, jessicaschedvin presents herself as a teacher and discloses that one of the marching Nazis had stolen her phone. Another influential tweet from that same user discusses what she labels as the ‘normalisation’ of the right-wing populist Sweden Democrats as she reports on witnessing representatives from that party and their youth organisation mingling with representatives from the conservative party youth organisation.

Much how like the 2014 event saw #almedalen activity from mainstream, established centre-right political actors dominating one of the identified clusters in Figure 5, the purple cluster found on the lower right-hand side of Figure 6 shows the influence of such users also during the 2018 Almedalen week. Individuals and parties affiliated with the mainstream centre-right in Swedish politics (cbildt, beckmansasikter, richardherrey, skyttedal, ericson_ubbhult, nya_moderaterna) make their mark here. While media personalities, journalists and media organisations (edward_blom, staffandopping, niklassvensson, expressen, omnipolitik) and financial professionals and interest groups (gunthermarder, foretagarna) make their presence felt as well, the section of Figure 6 just below the rather large node representing the user konsensuseliten (whose profile reads ‘to the right of you’) deserves particular scrutiny. Specifically, this section features a series of Sweden Democrat representatives and supporters (pontuspersson5, egorputilov) in addition to politicians and public figures even further to the right on the political spectrum (alfvonduring2, rogsahl, katjanouch, bardissimo, josefinutas, medborgsamling, afs_riks). Thus, while #almedalen activity previously saw clear right-wing influence, the 2018 event saw a split between more mainstream right-wing accounts and those representing more far-right views, thereby echoing developments in Swedish political life, to which we will return in the next section.

Finally, Figure 6 features two smaller clusters, demarcated by blue and orange colours, respectively. The blue cluster, visible in the upper right-hand corner of the figure, appears to be primarily related to health care issues. Here we find a series of users whose profiles suggest they have professional interests aligning with this theme: interest groups and trade unions (lakarforbundet, hjartlung, reumatiker), advisors and academics (ulfjanzon, hanswinberg, stefanjutterdal, cswahlin), media actors (dagensmedicin, kennedianskt) and tech companies (swedish_medtech, msdsverige, tietosverige). The orange-coloured cluster, then, is reminiscent of the green cluster identified for the 2014 event (see Figure 5) in that it features a series of policy actors primarily representing industry, as well as interest groups catering to the interests of such actors (energiforetagen, vattenfall_se, skogsindustrin, abbsverige, teknikforetagen, scaniasverige). As such, while non-party political policy actors made their presence felt during the 2018 event, #almedalen was clearly dominated by far-right political actors and events during this second-to-last year in our study. The 2010s in Sweden saw an increasing influence of far-right actors in mainstream social media, as has been evidenced by previous research (Åkerlund, 2020; Larsson, 2014).

While the first year of our study featured the hashtagged activity by comparably few users (N = 24,846) organised into one large cluster by means of retweet patterns, our final year of #almedalen activity features even fewer users (N = 7989) organised in three sparse but clearly defined structures, demonstrating the decreasing attractiveness of #almedalen as a space for agenda building among institutional policy actors. The green cluster, visible on the lower left-hand side of Figure 7, features a series of actors associated with right-wing political issues (richardherrey, beckmansasikter, josefinutas, ericson_ubbhult) as well as financial professionals and interest groups (gunthermarder, foretagarna). While many of these users are recognisable from previous years as detailed in Figures 4 to 6, the 2022 event also sees ledarsidorna_se enjoy considerable influence. Associated with what looks like the website of a hyperpartisan news provider which appears to be primarily operated by one person, ledarsidorna_se gains popularity within the #almedalen hashtag primarily by tweeting critical statements about the Social Democratic party. Thus, while this cluster also features a comparably large node associated with the environmental interest group ekomodernism, the bulk of users here is associated with right-wing political interests. By contrast, the blue cluster, found on the left-hand side of Figure 7, features a series of individuals (annaardin, helleklein, ulfbjereld) and parties (vansterpartiet, miljopartiet) associated with the left side of Swedish politics. Relatedly, the upper section of this particular cluster sees a series of policy actors with interests in international policy and foreign aid (sverigeivarlden, sida, utvmagasinet, nordensk).

Finally, the orange cluster, visible in the upper left-hand section of Figure 7, emerges as more of a mixed affair concerning the types of users. Here the most retweeted users are vjonaslundgren, a left-wing politician who primarily gains attention with a tweet on murders committed by right-wing extremists in recent Swedish history. He is joined by users associated with different trade unions (mternbo, ingenjorerna) to form this last cluster, which appears as less thematically coherent than the other clusters identified for the 2022 event – as well as for the preceding instances of #almedalen studied here.

Discussion

When studying the patterns over time emerging from actors connecting to each other through retweets using the #almedalen hashtag, a clear pattern emerges. In 2010, Twitter as a space for public discussion and agenda building still had a novel and unpolished character, creating opportunities for early adopters to engage in public discourse and exert influence. In step with the general adoption of journalists of Twitter as a news source (Djerf-Pierre, 2012), pioneer journalists start using the hashtag, as evidenced by the centrality of previously mentioned users such as matsknutson and emanuelkarlsten. Such adoption processes make it attractive for policy actors to use the hashtag as a space for providing information subsidies. The pioneer character during this early phase is also evident by the fact that there are no distinct clusters: the mix of journalists, politicians, communication professionals and non-institutional policy actors all seem to engage with – retweeting – each other.

As the 2010s progress, the activity in Almedalen as an event space (Figure 2), Swedish Twitter (Andersson et al., 2022) and #almedalen (Figure 3) increase. The peak year of the hashtag in 2014 is especially prescient considering that legacy media reporting was even higher in 2010 than in 2014 (Figure 1). The attractiveness of the hashtag as a space for agenda building is indicated by the influx of institutional, resource-rich policy actors and retweets becoming more common than original tweets (i.e. policy actors amplifying information subsidies, cf. Larsson, 2015). From previous studies of democracy festivals, there is a tendency for resource-rich policy actors to dominate (Sørensen and Lund, 2019; Wollebæk and Raknes, 2022). The results from 2014 indicate that in a situation where an online space is attractive to policy actors for providing information subsidies and for influencing the public discourse, this will attract resource-rich policy actors. In addition, the coherence of the hashtag in 2010 gives way to a more clustered network, in 2014 with political and issue-based groups centred around what is essentially the main character of that year’s Almedalen – the celebrity journalist niklassvensson.

Considering the success of #almedalen in 2014 and the stable levels of Twitter usage over the 2010s in the Swedish public sphere (Andersson et al., 2022), in combination with the fact that Almedalen as an event peaks in 2018 in terms of the number of events (Figure 2) and legacy media coverage (Figure 1), we might expect the patterns of 2014 to repeat in 2018. However, this is not the case: whereas Almedalen is flourishing, #almedalen attracts less interest than in 2014. At the same time, the 2018 retweet network provides evidence for the successful entry of right-wing and far-right actors into the hashtag (cf. Åkerlund, 2020; Larsson, 2014). Interestingly, we find centre-right actors of the traditional establishment in the same cluster as far-right actors – a finding that is tempting to interpret as a foreboding of the gradual rapprochement between Swedish centre-right parties and the far-right Sweden Democrats: initially a pariah upon entering Parliament in 2010, the Sweden Democrats – founded in 1988 by right-wing nationalists and neo-Nazis – gradually normalised in Swedish politics. After the 2022 election, they became the foundation of the incoming centre-right government. Their early exclusion led to a strong Internet presence, which they maintain while gaining access to legacy media (Nicolaisen, 2024).

Finally, in 2022, the interest in the Almedalen event has decreased substantially, and the attractiveness of the #almedalen hashtag seemingly collapsed. The result, as seen in Figure 7, is a map of dissonant subnetworks of centre-right, centre-left and issue network actors where there is little trace of the far-right dominance of 2018 or of large institutional policy actors from industry and communication professionals as seen in 2014.

From this summary, we can paint the trajectory of the #almedalen hashtag in four steps and understand it as the result of the changing attractiveness of the hashtag due to the attractiveness of Twitter, Almedalen, and Twitter hashtags as spaces for agenda building. This can help us understand which policy actors are dominant in the hashtag in different years: pioneers active in 2010 in one single network; establishment resource-rich actors in 2014; the concerted approach of the Swedish far-right to ‘take over’ established hashtags for exerting influence in 2018, and finally the lack of interest in 2022. Rather than unequivocally providing evidence for the dominance of resource-rich establishment policy actors found in previous research (Sørensen and Lund, 2019; Wollebæk and Raknes, 2022), or for non-establishment policy actors grabbing the opportunity of engaging in successful agenda building (Ihlen et al., 2022), we see that resource-rich actors are dominant when the attractiveness of the hashtag is at its peak (in 2014), but that the concerted effort of far-right non-institutional policy actors could achieve interpretative dominance in 2018. As noted, this chain of events can be seen as mirroring the successful mainstreaming of far-right discourse and political influence in Swedish society (Åkerlund, 2020). In addition, there is a clear tendency from more comprehensive networks connecting all users in 2010 and 2014, towards a clustering into separate retweet networks in 2018 and 2022, which points towards the development of a dissonant and fragmented public sphere (Pfetsch, 2018). These changes make it necessary to consider the obscure nature of the inner workings of Twitter. The year 2016 saw the platform change its algorithm to supposedly populate the user timeline with the most relevant tweets based on interests and activity as expressed on the platform (e.g. Oremus, 2017). Without being able to peek inside the metaphorical black box, the research community can do little except note that such changes are likely to influence empirical research. Given recent developments following the Musk takeover, at the time of writing it seems unlikely that such access will be granted. As Twitter has become X, new platforms like Bluesky have emerged. Their more open architecture may offer greater transparency in algorithmic content curation. While we focused on retweets due to their role in securing visibility and their popularity (see Figure 3), other interactions – such as mentions and replies – also likely influence tweet visibility. Future research could explore their roles in similar processes.

On a side note, the element of popularity in the social media logic (Van Dijck and Poell, 2013) means that single events, policy actors and even tweets can become viral to an extent that it makes a clear mark on retweet networks. We see such single instances of very popular content in the hashtag for three of the years, paradigmatically represented by a tech entrepreneur promoting ideas about intellectual property rights in 2010, a comedian challenging the leader of the Sweden Democrats in 2014 and the presence of Nazis in 2018 (but nothing in 2022).

Our findings suggest that agenda building in a hybrid media system is dynamic, shaped by platform-specific, event-specific, and societal factors. We highlight a complex transfer of issue salience across agendas, involving a diverse set of actors (cf. McCombs et al., 2014). This calls for more contextualised theoretical work in political communication and agenda-setting research. The results further suggest that there might be a strategic advantage in being an early adopter in a space, while resource-rich actors tend to have an advantage in crowded hashtags. In addition, actors can strategically work to ‘take over’ a hashtag at a stage when it still attracts significant attention but before it is perceived as irrelevant (cf. Åkerlund, 2020).

Thus, our findings point towards a direction in agenda-building research that considers both the platformisation of agenda building and strategic actions within agenda building in a dynamic hybrid media system (cf. Seethaler and Melischek, 2019).