Since Brexit and the election of Donald Trump, there has been much debate over how the spread of misinformation via social media is a danger to democracy. Some researchers have called fake news a serious threat to democracy: “The rise of fake news highlights the erosion of long-standing bulwarks against misinformation in the internet age. Concern over the problem is global” (Lazer et al., 2018: 1094). Yet, the same authors also say that “evaluations of the medium-to-long run impact on political behavior of exposure to fake news (for example, whether and how to vote) are essentially non-existent in the literature” (Lazer et al., 2018: 1095), though that situation is beginning to change quite quickly. This article examines the phenomenon of how digital media affect politics through the lens of the Swedish national election of 2018. It focuses on the role of dis- or misinformation, bots, and alternative news websites and how they supported the Sweden Democrats and their anti-immigrant agenda. The article draws out some lessons for how media and politics are changing in the world at large.
Like other Nordic countries – but unlike many countries in the world – Sweden has a strong tradition of public service media and a high level of trust in politicians and experts, and in the media in particular (discussed later). Moreover, the agenda is still set by public service and quality newspapers even if, as we shall see, many Swedes now often obtain their news from digital sources. Hence the question: has the new media landscape of online news and sharing of social media content affected even a country like Sweden, to the point of threatening the proper functioning of its democratic public sphere? Broadly, the answer will be no. But this answer should not lead to complacency or be taken to indicate that little has changed in the role of media in politics; in fact, there are specific changes – such as distrust in media among certain parts of the population, efforts to distort the news media, and a high-choice environment where nonmainstream political factions and their supporters can find online alternatives – that deserve closer scrutiny.
My approach in this article is theoretical as well as presenting empirical evidence about bots – I ask how the circumvention of gatekeepers can be theorised. To do this, mediatisation (Hjarvard, 2008), a theory that has been developed particularly strongly in Scandinavia, must be modified to cope with the new digital media environment. A review of mediatisation theory is beyond the scope of this paper, but the central thrust of the theory is that the media themselves have become increasingly important actors in their own right, establishing certain routines and formats in presenting news and political information that follow their own logic. This developmental logic, whereby media become more differentiated – although it has been extended to digital media (Finnemann, 2011) – needs to be extended still further: first, because digital media now increasingly target and tailor their messages to particular audiences; and second, because the attempt to do this in an underhanded and manipulative way (for instance, using bots), may undermine trust in the media system. This adds uncertainty about the media in their role as political actors.
I begin by providing background on the election with a brief account of the Swedish media system and the emergence of the Sweden Democrats. I then present evidence from the 2018 Swedish election, including analyses of Twitter, Facebook, and alternative news websites. In particular, I examine the role of bots on Twitter and how computational techniques could be used to gauge the spread of disinformation during the election. In conclusion, I place Sweden in the context of broader debates about the role of digital media in politics and argue that the uncertainty that has been created demands continued future vigilance.
Any account of the role of digital media must include how they operate differently from traditional media and fit into the media system as whole (Hallin & Mancini, 2004). Sweden belongs to the corporatist media systems of Northern Europe, where public service media play a strong role and where several parties compete in the political system. In Sweden, public service broadcast is still the main source of news (the following paragraphs draw on Schroeder 2019; see also Schroeder 2018: 67–70); however, the Sweden Democrats are supported by alternative media, news websites that promote a right-wing populist and antiimmigrant agenda. These alternative media have sizable audiences. According to the 2018 Reuters Digital News Report (Newman et. al., 2018), the following sites were used weekly in 2018: Fria Tider (11%), Nyheter Idag (10%), Ledarsidorna (8%), Samhällsnytt (8%), Nya Tider (6%), and Samtiden (6%). These are all “news” websites that support the Sweden Democrats and their anti-immigrant agenda; hence, the Reuters Report calls them partisan or alternative websites. The audiences of these sites may not seem large, but the largest offline source of news, with a 56 per cent audience share, is the Swedish public television broadcaster (SVT), and the alternative sites have an audience share that is not much smaller than other online sources like SVT News (15%) or the online versions of two main newspapers
The content on the websites of these alternative media portrays a Sweden that is quite a different picture from that portrayed by traditional media in Sweden – one in which immigrant crime is rampant and where multiculturalism, left-wing bias in public service media, Islam, and foreigners pose a threat to Swedish and Western culture (Holt, 2016). Immigration is also covered extensively in traditional media (see, e.g., Radio Sweden, 2017).
Immigration is also covered extensively in traditional media (see, e.g., Radio Sweden, 2017).
In Sweden, trust in the media – and also in politicians and experts – has remained steadily and comparatively high since the 1970s. One exception is that in very recent years – and in the wake of the “migration crisis” in particular, and primarily among Sweden Democrat supporters – this trust has declined (Sannerstedt, 2017; Andersson & Weibull, 2018). At the same time, specifically on the issue of immigration, the public has not generally trusted the media: among Sweden Democrat supporters, distrust on this issue stood at 93 per cent, while it was 60 per cent among the general population (Rydgren & van der Meiden, 2019). In Sweden, too, there has been an increasingly antagonistic atmosphere between populists and the media, as Herkman (2018) has shown by reference to the recent increase in the number of media scandals relating to the Sweden Democrats. The scandals have caused outrage in the media and among the public, yet they also enhanced the visibility of the Sweden Democrats. And among the supporters of the Sweden Democrats, Herkman (2018: 351) says, “the media hunt [against the Sweden Democrats] represents a corrupt elite and their excessively liberal opinions”. It can also be noted that the Sweden Democrats have, partly due to these scandals, received mainly negative coverage in traditional media (Rydgren & van der Meiden, 2019; see also Oscarsson & Strömbäck, 2019).
The Sweden Democrats are a populist party that regards itself as upholding Sweden's main political tradition – the folkhem [people's home], a strong welfare for all – though their message in the media and campaign material (mostly between the lines) is that citizenship rights or benefits should be restricted to “native” Swedes and exclude “others”. In Sweden, alternative media have also attacked “multicultural elites” and their “political correctness”. The politics of the Sweden Democrats is mainly focused on immigration, and their pressure has led to the implementation of more stringent immigration policies since 2016. The initiative for this change came from the Social Democrats on the left, but they were clearly responding to pressure from the Sweden Democrats. It is not clear how popular the new restrictive policy is, but it has outflanked the Sweden Democrats on their main issue.
In the television debates among the party leaders in the run-up to the 2018 election, debates about immigration and multiculturalism played a major role. Consider the controversy over the role of the public service media in their coverage of the Sweden Democrats’ immigration stance (The Local, 2018).
Consider the controversy over the role of the public service media in their coverage of the Sweden Democrats’ immigration stance (The Local, 2018).
Support for the Sweden Democrats has been steadily rising in recent Swedish elections. It gained 5.7 per cent of votes, or 20 parliamentary seats, in 2010, 12.9 per cent and 49 seats in 2014, and 17.5 per cent and 62 seats in 2018. This means that they were the third-largest party behind the left-wing Social Democrats and the right-wing Conservatives [Moderaterna] –putting the Sweden Democrats into a potential “kingmaker” position whereby the party could dictate alliances. But both elections in 2014 and 2018 led to coalition deals whereby the Sweden Democrats were excluded from being part of the government. Minority governments were formed on both occasions with great difficulty and after protracted negotiations. After the 2018 election, it took six months to form a minority government. And being “locked out”, the Sweden Democrats have been able to claim, in typical populist fashion, that “the people” are not being represented because a “deal” has been made “behind closed doors” (though there has been continual speculation in the media since at least 2014 about whether the other parties would abandon the
Attacks on the media, how the media – especially digital media – have become distorted by mis- or disinformation, and how civility in public discourse has declined have not become issues for the Swedish media to the same extent as in the US (though the Swedish media often report on American politics along these lines with the implicit warning that the same could one day happen in Sweden). The Swedish government and the media also took several steps during the 2018 election to counteract foreign meddling in elections (including the role of bots, see below), made fact-checking efforts,
The election in 2018 took place against the background of extensive public discussion about online disinformation, and extensive measures were taken to counteract malevolent actors. Here, we can leave to one side the debate about how to define disinformation and misinformation (but see Woolley & Howard, 2018; Jack, 2017) and focus on the distortion (which includes both mis- and disinformation) of political information by means of bots; distortion thus consists of efforts to underhandedly skew information. As part of the effort to counteract the distortion of political information during the Swedish election, the Swedish government, via the Swedish Civil Contingencies Agency (MSB), tasked the Swedish Defense Research Agency (FOI) with monitoring digital media during the Swedish election. The FOI undertook a study (in which I took part) of bots on Twitter in the run-up to the election. The study collected election-relevant tweets over a period of more than six months before and just after the election. Using machine learning techniques, the study revealed a rapid increase in the number of election- and politics-related tweets, especially in the month before the election. The study also found certain skewed patterns in criticisms of parties and of links to alternative or partisan websites. Some of the details of the study are now worth presenting in greater depth.
Before we do so, however, it can be mentioned that there were two other studies that specifically examined disinformation in the Swedish election. The first report, by Colliver and colleagues (2018), monitored a number of social media and extremist websites in the run-up to and shortly after the Swedish election. They found that these were mainly aimed at influencing the perceptions of international audiences rather than influencing the election directly. Importantly, the authors found no bot activity or amplification activities originating from the Kremlin. The authors of the report also detected many tweets shortly after the election with the theme of valfusk [election cheating] (Colliver et al., 2018). The second report, by the Computational Propaganda project at the Oxford Internet Institute, analysed the Swedish election (Hedman et al., 2018) as part of a series on propaganda in major elections. The report on the Swedish election focused on Twitter and the spread of junk news, which was defined by using several criteria in contradistinction to professional news. The study found that on Twitter, “for every two links of professional news content shared Swedish users shared one junk news story – with 22% of all URLs shared, this was the largest proportion of junk news across all the European elections we have studied”, though it was similar to the level during the 2016 American election, and that “most of the junk news originates from Swedish outlets” (Hedman et al., 2018: 5).
What role did bots play in the election according to the FOI study? Bots typically refers to computer programs developed to carry out tasks that would be highly demanding or time-consuming to carry out otherwise. It is worth noting that while bots are often used for malicious purposes, they can also be used for benign purposes, such as updating Wikipedia entries. In the FOI research (Fernqvist et al., 2018) on bots in the Swedish election, computational techniques were used that do not distinguish whether a bot is a software program or a person who frequently and continuously posts and retweets content in a bot-like manner. The effects of these two behaviours is the same regardless of whether software or a human is responsible for the account. Technically, therefore, the term automated behaviour should be used instead of bot, but since bot is commonly used, bots is used in what follows.
In the period from 5 March to 30 September 2018, tweets containing keywords concerning the election and Swedish politics were collected – for example, #valet2018 or #svpol – from the Twitter streaming API. The technical details of the method for detecting bots as opposed to genuine accounts (and suspended or deleted accounts) will be left to one side here – they are described elsewhere (Fernqvist et al., 2018). The collection yielded 1,005,276 tweets from 70,973 accounts, and showed that 8 per cent of the content and 6 per cent of the accounts that tweet about Swedish politics and the elections were bots (and this excludes accounts, for example, that merely update news websites). The figures would be somewhat higher if we were to assume that accounts suspended by Twitter – which Twitter often does for a variety of reasons – are also bots.
The study also examined the top links that were contained in the tweets, and it found that both among bots and genuine accounts, among the top ten most linked websites (including SVT and YouTube), there were several alternative or partisan websites such as Samhällsnytt and Nyheter Idag (Nordfront, an extremist website, is among the suspended Twitter accounts). But to get a better sense of the kind of content spread by bots, the study also analysed a random sample of 935 tweets (which was deemed a sufficient sample) spread during the week before and just after the election. The tweets were put into eight categories, such as “media criticism”, “criticism of elites”, and the like, including two with subcategories for different “parties supported” or “parties criticised” (and including “unknown” or uncategorisable). This categorisation was arrived at inductively by examining a subsample and identifying the most relevant categories. The tweets were then all coded by two researchers independently, with an acceptable degree (0.61 Fleiss Kappa) of inter-coder reliability. The study compared the content for both genuine accounts and bot accounts, which did not reveal large differences except that bots communicate slightly more messages that are critical towards immigration, Islam, media, and political parties.
The study also coded these tweets according to those that expressed support and criticism of different parties, coded in a binary way (though many did not express support or criticism), and found that both bots and genuine accounts expressed a disproportionate amount (relative to the election results) of support for AfS (Alternative for Sweden) party – a newly formed party to the right of the Sweden Democrats that received only 0.3 per cent of the votes but received 29 per cent support from bots and 27 per cent from genuine accounts. The Sweden Democrats (SD) received 47 per cent support from bots and 28 per cent from genuine accounts – again, disproportionate to the election result (17.5%). All other parties, in comparison, received less than 10 per cent of support from both types of accounts. When we look at criticism of parties as opposed to support, it is the Social Democrats that stand out, receiving more than 40 per cent of criticism from bots and genuine accounts, while other parties receive at most 15 per cent, but typically well under 10 per cent of criticism.
What we see is a distorted picture of Swedish politics on Twitter. It is clear that bots tweeting about Swedish politics and the 2018 election tended to spread immigration-critical and populist-supporting messages, at least for the hashtags related to the election and politics included in the FOI study. One limitation of the FOI analysis was that it could not determine where the bot activity originated (though the other two studies concluded they were not primarily from abroad). Regardless, these bots created a climate of uncertainty about the role of disinformation in the election, which meant that the public was also uncertain about how social media were shaping the election. The efforts of the FOI and other studies – which were widely covered by television, radio, and newspapers – helped to raise awareness. Yet the problem of malicious influence during elections is likely to persist. Twitter and Facebook, among others, have recently begun to take active measures against bots and disinformation by means of self-regulation and other efforts at regulation are also under way, though whether these are effective remains to be seen.
Changes to the media landscape have become entrenched whereby in Sweden, alternative news media provide the populist party and its supporters with an opportunity to continually challenge the government and traditional media, including during elections. But online alternatives and efforts to distort political information must be put into context. Oscarsson and Strömbäck (2019) produced an authoritative guide to media uses during the 2018 election. There is not much to disagree with, except that they underplay the role of digital or non-traditional media: they challenge the “conventional wisdom [of the] large potential effects of on-line activities per se” because they don’t find much use for “‘alternative’ media [with] more or less explicit and party-affiliated agendas [and an] overall low magnitude of social media use” during the election campaign (Oscarsson & Strömbäck, 2019: 335). But that is curious, because they show that more people use online news sources than traditional media, and they do not show figures for the many people that use alternative websites, as the Reuters report, cited earlier, does (it can be noted in passing that, according to the same report, Denmark, Finland, and Norway also have such anti-immigrant alternative websites, though with smaller audience reach). We do not know, for example, whether the 6–11 per cent of Swedes who looked at alternative or partisan Sweden Democrat supporting websites were able to support their views from these websites. Nor do Oscarsson and Strömbäck (2019) discuss the considerable debate about the role of bots and other sources of online disinformation that featured prominently in the weeks just before the election: even if the media perhaps exaggerated this role, the uncertainty media reports helped to create about the role of digital media will itself have influenced voters, perhaps making them uncertain about the media and the democratic process generally.
We also know little about which websites were shared on social media. When Sandberg and Ihlebæk (2019) compared shares of ten “right-wing alternative media” with “traditional news media” in 850 public Facebook groups collected from three weeks before and one week after the election, they found that 28 per cent of the shared content came from these ten alternative media, much of it (41%) concerned with the topic of immigration. This measurement is limited, as the authors note, since it is not known how much content was shared on Facebook overall, nor what was shared privately. Still, it shows a great deal of engagement with right-wing alternative media content. And although the authors did not find content that was “fabricated”, it found that the content, both produced by the sites themselves and taken from other media, was “recontextualized to make a political statement” (Sandberg & Ihlebæk, 2019: 435). In other words, those who are not happy with the content of traditional news media can reshape it towards theIr own proclivities.
The idea of mediatisation – how different media formats shape the political process – therefore also needs to be extended to take into account the various alternative online channels, or channels that do not follow established journalistic norms and that fall outside of traditional media institutions and also outside of the conventional ways the public engages with political information. To be sure, the agenda is still set by traditional media, but for those who are unhappy with this agenda, there are plenty of alternatives becoming entrenched as new forms of mediatisation, and with new formats, such as being mainly critical of or presenting an alternative to traditional media. And for the wider public too, the media environment has been undermined in the sense that bots also make political information more uncertain. If this new uncertainty about online alternative news and bots, via the coverage of social media in elections and politics, leads to greater uncertainty about the media as a political actor, then perhaps the role of traditional or professional media now needs to highlight how these professional and traditional media set themselves apart from the “unruly” online world. This could include emphasising objectivity, impartiality, and inclusiveness, in contrast to other, non-gatekept sources. In other words, the new element that is added to mediatisation (distortion or disturbance of a reliable media environment by bots and alternative news, and also users sharing content on Facebook) needs to be explicitly and critically responded to. This response could include regulating bots, but also, insofar as alternative news is not illegal, by professional news media distinguishing their own standards and formats from online sources where those standards and formats are not being met. In other words, traditional professional media (also in digital format) need to reassert their role as neutral arbiters in the face of the unruliness of online political information, most of which is legitimate free expression (even if part of it will also need to be regulated, but regulation will also be contested and attempts will be made to circumvent regulations, so media will have to assert themselves in the face of these efforts too!).
What we can see on Twitter and on alternative websites and other social media is that online political activity, including content shared, posted, and viewed, is in part quite different from the content of traditional or professional media. We cannot be certain how many people accessed this content, but we know that it offers populists like the Sweden Democrats and their supporters an alternate source for views unlike those provided in traditional or professional media. The conclusion is therefore ambiguous, but perhaps productively so: Yes, even in Sweden some digital media have played a new and harmful role, if by harmful we mean media that do not subscribe to the role of news media whereby they should be objective, impartial, and presenting a broad, diverse, and inclusive picture of society. And yes, the role of deliberate or underhanded disinformation needs to be monitored and counteracted. Yet, there will also continue to be ways to circumvent the gatekeepers of traditional and professional media that allow right-wing populists and others to get their message across, mostly by legal means. So, another conclusion (which is inconclusive) is that no, digital media did not, as far as we know, play a decisive role in the recent Swedish election (or in the 2016 American election, see Schroeder, 2019). Online media therefore represent a disruption – some of which (deceitful disinformation) needs regulation, and some of which (alternative news that does not follow journalistic standards) deserves to be counteracted – that Sweden and other democracies must continue to address and live with. This disruption will also require, as a task for future research, a rethinking of mediatisation which includes this unruly or distorting role of online information, both in its own right and in making the political information environment more uncertain – that will also make it possible to establish how the properly functioning role of media could be reasserted. These dysfunctions would not constitute a problem for Swedish democracy if there were no politicians and their supporters who felt left out by traditional media. But then, they would not be able to continue successfully challenging the political status quo, as they have already done, without online media.