The virus of polarization: online debates about Covid-19 in Germany

ABSTRACT

To what extent do online debates display features of political polarization and in how far does polarization pose a problem for democracy? We zoom in on affective polarization: the formation of societal groups with hostile feelings towards each other, arguing that affective polarization is particularly problematic for democracy if it features elements of political intolerance, which undermines key tenets of even the most conflict-prone theories of democracy. While affective polarization has been on the rise in several countries, Germany has been considered to be a country with low, and even declining levels of affective polarization. But does this still hold true during the Covid-19 pandemic, which saw a rapid rise in conspiracy theories? Based on a qualitative discourse analysis of online debates about Covid-19 on the Facebook platforms of a mainstream and a non-mainstream German media outlet, we find strong traces of affective polarization on both platforms, involving clear indications of political intolerance. Our findings suggest that the democratic discourse is threatened by the nature of online debates about Covid-19, and it is threatened not only by anti-rationalist conspiracists at the ideological extremes but also by the intolerance of more moderate rationalists at the centre of the political spectrum.

KEYWORDS:

• Polarization
• conspiracy theories
• democracy
• social media
• public debates

Introduction

The Covid-19 pandemic has had an unprecedented impact on many countries worldwide. Starting in Wuhan in December 2019, the Sars-CoV-2 virus has spread across the globe, taking a death toll of several million. Countries around the world introduced far-reaching measures to contain the spreading of the virus, ranging from school closures, travel restrictions and bans on public events to local curfews and national lockdowns. In response to these unprecedented events, Covid-19 also dominated media attention. Especially during the pandemic’s lockdown months, when citizens were confined to their homes and had no physical contact with many people with whom they normally interact, this physical confinement also triggered widespread online debates on social networking sites, such as Facebook and Twitter, to discuss the pandemic and the measures taken.

Yet, many social media debates on Covid-19 revolved around conspiracy theories worldwide. While some of these theories deny the pure existence of the virus, for example by stating that Covid-19 was an invention by the powerful elite to control the people, others have emerged around different causes of the rapid spread of Covid-19, such as the idea that the virus was man-made and had escaped from a laboratory near Wuhan, China, or that the spread of the virus was connected to the rollout of 5G mobile network technology.

The development of conspiracy theories in societal crisis situations is not a new phenomenon (see Van Prooijen and Douglas 2017). Prior research has shown that situations of high uncertainty can be the ideal breeding ground for the spread of misinformation and thus the development of conspiracy theories of all kinds (Sunstein and Vermeule 2009; Starbird et al. 2014; Bessi et al. 2015; Del Vicario et al. 2016; Van Prooijen and Douglas 2017). The Covid-19 pandemic represents such a situation of high uncertainty. Especially when Covid-19 appeared and started to diffuse, medical experts had very little in-depth knowledge about where the virus came from, how it is transmitted between people and how dangerous it is. According to Van Prooijen and Douglas (2017), when negative and unexpected events occur, such crises often trigger sense-making narratives among citizens that become part of their historical narratives. Conspiracy theories thus develop around a certain explanatory belief in response to a situation of high uncertainty as an attempt to give meaning to the situation (Bruckmüller et al. 2017).

One of the main problems with conspiracy theories is that they can lead to, or aggravate, political polarization. Conspiracy theories are essentially based on the self-contained quality of their arguments combined with the spread of mistrust in conventional knowledge-producing institutions (Hofstadter 1965; Sunstein and Vermeule 2009). This leads to extreme resistance to correction and a dualistic worldview with a strong separation between us (the people) and them (the power elite). Conspiracy theories therefore not only produce political divides, but their emergence and proliferation are also stimulated by pre-existing political divisions, suggesting a vicious cycle that is likely to deepen political divides more and more.

Growing political division can pose a serious threat to democracy, especially when conspiracy theories promote polarization that exacerbates not only preference-based political division (hereafter referred to as ‘ideological polarization’) but also divide citizens into different political camps that begin to develop hostile feelings towards each other (hereafter referred to as ‘affective polarization’; Iyengar, Sood, and Lelkes 2012). Not all negative feelings towards political opponents are a threat to democracy, though. We propose a two-stage conceptualization of affective polarization. Affective polarization starts with the emergence of hostile feelings towards opponents based on group identities (Stage 1). As a result of ever-growing negative outgroup identities, hostile groups may then also develop outright political intolerance towards the political opponent (Stage 2). In many cases, the full-fledged version of affective polarization at Stage 2 is connected to anti-elitist narratives, the binary division of good and evil, and the constitution of these two groups as internally homogenous (Vaughan and Heft 2022). We argue that Stage 1 affective polarization is still compatible with more conflict-oriented approaches to democracy whereas Stage 2 affective polarization poses a threat to democracy even for the most conflict-oriented theories of democracy. If the members of the hostile political camps stop accepting each other as legitimate participants in democratic exchange and instead start to regard each other as morally evil enemies, democracy is seriously threatened.

Previous studies show that social networking sites not only easily amplify and spread rumours and questionable information (Cinelli et al. 2020), but they can also facilitate the creation of new conspiracy theories that thus feed political polarization (Grömping 2014). This conclusion was confirmed by a recent study on the spread of fabricated news and conspiracy theories on the Facebook pages of German alternative media outlets (Boberg et al. 2020). There have also been first studies on the impact of social media use on affective polarization (Lelkes, Sood, and Iyengar 2017; Suhay, Bello-Pardo, and Maurer 2018; Asimovic et al. 2021) and on the prevalence of affective polarization in social media debates (Harel, Jameson, and Maoz 2020; Yarchi et al. 2021). Yet to this point, little is known about the extent to which online debates not only display features of identity-based affective polarization but also show signs of outright intolerance vis-à-vis the political opponent. The exceptionally threatening global crisis caused by the Covid-19 pandemic provides us with an ideal opportunity for studying the nature of societal debates on social networking sites. By analysing the online social discourses on Covid-19 in-depth, we are able to gain a better understanding of how people from different backgrounds interact with each other, to what extent they are prone to spreading misinformation and conspiracy theories, and how they react to people who make claims that are opposed to their own worldviews.

To capture the nature of online societal discourse, we analyse the case of Germany – a country that has previously displayed low, and even declining, levels of affective polarization (Boxell, Gentzkow, and Shapiro 2020). For our study, we choose a qualitative study because it allows for an in-depth analysis of public discourses to learn more about the role of outright political intolerance in people’s affective attitudes toward their opponents. We conduct a discourse analysis of the debates about Covid-19 on the Facebook platforms of two different German media outlets: the mainstream, liberal platform Süddeutsche Zeitung (SZ) and the non-mainstream, conspiracy-disseminating platform Russia Today Germany (RT). Our choice of these two media outlets was based on the expectation that affective polarization would be present in heterogeneous ways on both platforms. Such differences may occur with regard to the homogeneity or heterogeneity of expressed views, the endogenous or exogenous nature of polarization, i.e. polarization between commenters or a platform or polarization directed towards outside groups, or perhaps the use of populist tropes. Our analysis of 1,104 Facebook comments shows that the debates on both media outlets exhibit strong traces of affective polarization involving not only identity-based hostile emotions but also clear signs of mutual political intolerance. In this article, we argue that if this type of discourse were to prevail on other social media platforms, on other issues, and in other countries, it would form a major challenge to democracy.

Political polarization, public debates, and democratic theory

Political polarization is a process marked by ‘increasingly harsh divides between opposing political camps and diminishing shared political ground’ between these camps (Carothers and O’Donohue 2019, 1). From the literature, we can distinguish two dimensions of political polarization: ideological and affective polarization (Iyengar, Sood, and Lelkes 2012). As for the former, ideological polarization exists when political parties and citizens have widely diverging preferences over salient political issues. This is the minimal definition of polarization, and many empirical contributions focus on this dimension alone (Abramowitz and Saunders 2008; DiMaggio, Evans, and Bryson 1996). In contrast, affective polarization aggravates the preference-based divisions resulting from the first dimension. It exists when parties or citizens that are split into different political camps start to develop hostile feelings towards each other (Iyengar, Sood, and Lelkes 2012). If preference-based divisions become deeper and deeper, opposing camps are likely to develop group-based identities, and these identities then give rise to positive feelings towards members of the ingroup and negative emotions towards members of the outgroup. This affective dimension of polarization can easily spill over into political intolerance towards the respective other camp: political disagreement then turns into a war between groups that have ceased to accept each other’s positions as legitimate but consider the other side as morally ‘evil’ (Iyengar and Westwood 2015; Vaughan and Heft 2022).

Previous research on affective polarization has focused strongly on the US, where there is a clear trend towards rising levels of affective polarization between Democrats and Republicans (Lelkes 2016; Iyengar et al. 2019; Boxell, Gentzkow, and Shapiro 2020). Likewise, affective polarization has been shown to be on the rise in countries like Switzerland, France, and Denmark, while countries such as Norway, Sweden and Germany have seen a decline in affective polarization (Boxell, Gentzkow, and Shapiro 2020).

But why should we be worried about affective polarization? Different theoretical approaches to democracy provide different answers to this question. In the following section, we first develop a two-stage conceptualization of affective polarization: polarization based on group identities (Stage 1) and polarization additionally involving political intolerance towards opponents (Stage 2). We then briefly discuss three approaches to democratic theory that provide widely diverging answers to the question of whether ideological and the two different stages of affective polarization are compatible with democratic exchange: deliberative democracy, pluralist democracy, and agonistic democracy. Each of these approaches has a different view on the benefits and perils of political polarization. Ultimately, however, even the most conflict-prone approach, agonistic democracy, considers affective polarization that turns into political intolerance a threat to peaceful democratic exchange.

A two-stage conceptualization of affective polarization

A close reading of the literature on affective polarization suggests that we should conceptualize affective polarization as a two-stage phenomenon. The first stage is rooted in group identities: hostile outgroup identities. It is the result of a radical kind of ideological polarization, where ideology fundamentally shapes individuals’ social identity and affiliation with a group (Iyengar et al. 2019). Group affiliation is a deeply rooted social desire and has the purpose of defining one’s own identity to simultaneously take up a position in society. By associating with an ideology, individuals become part of a group and thereby accept its attitudes. In radical cases of social conflict with another group, this ingroup identity can turn into a hostile outgroup identity, involving an emotionally charged us-vs-them attitude which rejects everything coming from the outgroup as being incompatible with the identity of one’s own group.

This phenomenon forms the basis for the second stage of affective polarization, which involves political intolerance. The opposing camp is then seen as an existential threat to one’s own group as well as to society as a whole (Doherty, Kiley, and Jameson 2016; Garcia-Guadilla and Mallen 2019), and is supposed to not only be wrong but inferior, illogical and morally illegitimate (Hacker et al. 2006; Brasted 2012; Abramowitz 2013). As a consequence, finding compromises with the opposing side, or even accepting to be governed by the opponents, becomes increasingly detested (Mutz 2006; Sunstein 2007; Mutz and Young 2011). In essence, such expressions of intergroup hostility, moral depreciation, and uncivil behaviour mean that opposing sides cease to accept each other as legitimate participants in the democratic process.

Three democratic views on political polarization

How do different approaches to democratic theory perceive the different dimensions and stages of political polarization: ideological polarization, affective polarization based on hostile outgroup identities, and full-fledged affective polarization also involving political intolerance? We discuss three approaches that cover a wide spectrum of perspectives on political polarization: deliberative, pluralist and agonistic democracy.

Proponents of deliberative democracy set the tightest limits to polarization. Deliberative democracy builds on the idea that normatively acceptable democratic decisions are best reached by public deliberation: citizens engage in truth-oriented exchanges of views and seek to persuade those with other views by presenting good arguments. Ultimately, the goal of deliberation is to reach consensus: a solution that is supported by, or at least acceptable to, everybody affected (Habermas 1998, 239–252; see also Dryzek 2009; Fishkin 2011). While different supporters of deliberative democracy vary in the extent to which they consider gradual deviations from the ideal of a rational discourse acceptable (Bächtiger et al. 2010), it is undisputed that political polarization is a major problem for discursive models of democracy. This is true for both dimensions of polarization. Ideological divisions may to a certain extent be overcome by exchanging rational arguments, but reaching consensus becomes increasingly unlikely as ideological divides deepen. And affective polarization undermines the very basis of deliberation: if people consider each other evil, there is no basis for engaging in truth-oriented rational deliberation in the first place (Strickler 2018).

Supporters of pluralist conceptions of democracy, perhaps the most widely-accepted model of democracy, have fewer problems with a certain degree of ideological division, but they also consider inter-group hostility a major threat to democracy. Pluralist democracy rests on the idea that citizens have different political views and interests. To make themselves heard in the political process, like-minded citizens organize in interest groups and political parties. The goal of the democratic process is that all of these different political interests can voice their concerns unrestrictedly and equally and that they can find, through an open and fair democratic exchange, compromises that are acceptable for many or most of them (Dahl 1971, 1998). Pluralist democracy is largely compatible with ideological polarization. That people have different political convictions belongs to the central tenets of pluralism, which means that, in principle, such divisions are in line with the tenets of pluralist democracy. If the divisions within society become so deep that they turn the pluralist exchange of views into disagreement in principle, however, pluralist democracy runs into trouble. Dahl (1998, 149) thus considers a social structure ‘that is culturally fairly homogenous’ as a precondition of democracy. Cultural conflicts, forming around religious, ethnic, or moral issues in his view often lead to political demands that are considered a matter of principle. Such ‘non-negotiable’ demands thwart the process of finding mutually acceptable compromises (Dahl 1998, 150). Affective polarization is an even bigger problem for pluralist conceptions of democracy. A certain amount of negative emotions towards political competitors might be compatible with pluralist conceptions of political exchange, but intergroup hostility rooted in deeply-held group identities runs counter to the idea that groups should be able to reach compromises. Political intolerance, finally, is entirely at odds with pluralism. An essential condition of pluralist democracy is that all groups and parties accept each other as legitimate participants in the political process (Herman 2017). If opponents are seen as morally ‘evil’ enemies, this key condition is violated. Under these circumstances, fair negotiations leading to mutually acceptable compromises cannot take place.

The theory of agonistic democracy, finally, is the approach that tolerates the largest degree of polarization. It considers ideological polarization not only unproblematic but the very essence of democracy. Agonistic democracy insists that political decision-making is fuelled by competing ideological worldviews. At the heart of democracy is thus conflict between these competing ideological camps (Connolly 1991; Mouffe 1999, 2013). This emphasis on conflict and contestation sets agonistic democracy apart from deliberative democracy, which is geared towards consensus-seeking, and from pluralist democracy, with its focus on political compromise. And it makes it the only of our three theories of democracy that is compatible with far-reaching ideological polarization. Agonistic democracy is also in line with the first stage of affective polarization. Deep ideological divisions may easily lead to strong feelings of identification with one’s ingroup and a negative identification with one’s outgroup(s). Like deliberative and pluralist democracy, however, agonistic democracy is fundamentally challenged by the second, affective, stage of polarization, where negative outgroup identities spill over into political intolerance. Agonistic conflict presupposes that members of different political camps accept each other as ‘adversaries who recognize the legitimacy of the demands of their opponent’ (Mouffe 2013, 138). This sets the agonistic model apart from ‘antagonism’, which is marked by hostile interactions between enemies.

In sum, political polarization poses major problems to all three models of democracy. But the assessment of polarization varies between the two dimensions of polarization, and between the two stages of affective polarization (Table 1). Ideological polarization is at least partly in line with all three models of democracy. The consensus-orientation of deliberative democracy makes it most vulnerable to major ideological divides, but a certain amount of ideological polarization can be overcome in deliberative processes. Pluralist democracy can largely be reconciled with ideological polarization, but if ideological differences become matters of principle, they can obstruct the process of carving out political compromises. Agonistic democracy, finally, is fully in line with the idea of ideological polarization. Affective polarization, in contrast, poses a problem to all three models of democracy, especially at the second stage, where negative outgroup identities spill over into political intolerance. For the most conflict-oriented approach, agonistic democracy, negative emotions vis-à-vis political opponents rooted in identity conceptions may be unavoidable as a side-effect of ideological polarization. But even proponents of agonistic democracy insist that all parties to the democratic process need to accept each other as opponents with legitimate political goals. If opponents consider each other morally evil enemies, not even agonistic democracy can be sustained. The same is even more true for pluralist democracy and deliberative democracy.

Table 1. Compatibility of ideological and affective polarization with three models of democracy.

	Ideological polarization	Affective polarization
		Hostile outgroup identities (Stage 1)	Political intolerance (Stage 2)
Deliberative democracy	Largely incompatible	Incompatible	Incompatible
Pluralist democracy	Largely compatible	Partly compatible	Incompatible
Agonistic democracy	Compatible	Compatible	Incompatible

Case selection, data and methods

The aim of the study is to investigate the presence of affective polarization within the public debate developing around the Covid-19 pandemic and to identify whether this affective polarization involves elements of intolerance towards the opponent so that it becomes a potential threat to democracy. We opt for a qualitative approach as this facilitates in-depth analysis of public discourses in order to learn more about the role of outright political intolerance in people’s affective views vis-à-vis their opponents. Such a fine-grained analysis is hard to achieve with quantitative approaches to text analysis, whose strength lies more in the classification of large volumes of text into relatively broad categories such as emotional/non-emotional or positive/negative emotional sentiment (see, for example, Yarchi et al. 2021).

In order to turn this research interest into a manageable empirical project, we needed to select a suitable country setting in which to study these debates, a social networking platform that allows us to identify forms of affective polarization in debates, and concrete sites on this platform on which we could search for evidence of different facets of affective polarization.

Regarding the country context, we decided to focus on societal debates in Germany. This was not only driven by pragmatic considerations of language proficiency among the authors but, more importantly, also by the idea that Germany is among the least-likely cases of political polarization. In contrast to a country like the US, for which previous research has already uncovered important traces of both ideological and affective polarization (Abramowitz and Saunders 2008; Hetherington and Weiler 2009; Lelkes 2016; Iyengar et al. 2019; Boxell, Gentzkow, and Shapiro 2020), Germany still appears to be relatively unpolarized. Even though the country now also has a populist radical right party, the Alternative for Germany, the establishment of this party happened much later than in many other countries, and its electoral support base is still comparatively low. Germany is marked by a stable centrist coalition government, comparatively high levels of public trust in government (OECD 2020, 238),¹ and above-average levels of trust in the media (Newman et al. 2019, 87). In previous studies, the country has also displayed low, and even declining, levels of affective polarization (Boxell, Gentzkow, and Shapiro 2020). If we can identify patterns of affective polarization in German debates, therefore, we can expect to find even more of these patterns in countries with less benevolent political cultures.

Concerning social networking platforms, we decided to select Facebook due to its text focus and the fact that it does not limit the length of postings. Twitter, the other major social media platform primarily focused on text exchange, in contrast, sets a tight limit for the length of individual messages. Facebook thus provides a more favourable environment for meaningful debates with longer contributions. Moreover, the number of people in Germany who use Facebook is about three times as high as the number of Twitter users.² The official Facebook sites of several news media regularly post their articles online and consequently trigger a vivid discussion, especially on divisive issues such as Covid-19. Although people who write user comments on online news platforms are driven by very specific motives (Springer, Engelmann, and Pfaffinger 2015), suggesting that the views they express should not be interpreted as generalizable, many more people actually read these comments, which means that the impact of the opinions voiced in such user comments may be much higher than just representing the views of a specific minority of active commenters. Since most previous studies on affective polarization have relied on surveys (Lelkes 2016; Lelkes, Sood, and Iyengar 2017; Suhay, Bello-Pardo, and Maurer 2018; Iyengar et al. 2019; Boxell, Gentzkow, and Shapiro 2020; Asimovic et al. 2021; but see Harel, Jameson, and Maoz 2020; Yarchi et al. 2021) analysing Facebook debates has the additional advantage that it may provide insights into the prevalence of affective polarization in people’s political behaviour, as opposed to their political attitudes.

With a view to the concrete Facebook sites analyzed, we started from the observation that there are significant discrepancies between the readerships of different news sites regarding their ‘demographic profiles, political orientations and their perceptions of certain socially charged issues, such as immigration’ (Godwin et al. 2018, 851). In order to maximize the diversity of users, and hence the viewpoints they expressed, we selected one mainstream news medium, the newspaper Süddeutsche Zeitung (SZ), and one non-mainstream news medium, RT Deutsch Production, for this study.

With a reach of more than 1.25 million readers, Süddeutsche Zeitung is one of the most widely read daily newspapers in Germany. Next to the readers of its daily print issues, its articles also reach around 770,000 subscribed users on Facebook.³ According to its editorial statutes, the SZ strives for ‘liberal, democratic forms of society based on liberal-social principles’ and is perceived by the public as left-liberal journalism.⁴ The SZ is classified as a leading German quality news outlet that shapes and influences social communication as well as the public (Fengler and Vestring 2008).

RT Deutsch Production (around 488,000 followers as of 06/2020) is a subsidiary of the state-owned Russian media company Russia Today, which has been shown to spread conspiracy theories in order to legitimize the Russian government’s foreign policy (Yablokov 2015). As stated in its imprint, the German branch of RT sees itself as a counterbalance to the ‘biased and often interest-driven media mainstream’.⁵ RT specifically broadcasts to audiences with pre-existing anti-establishment beliefs (Elswah and Howard 2020, 642). It skilfully picks up on existing doubts within German society and instrumentalizes them for its own purposes (Spahn 2018). According to the Federal Office for the Protection of the Constitution, ‘state companies are disguised as apparently independent media in order to disguise their affiliation to the Russian state and to subtly influence public opinion’ (Bundesministerium des Innern 2019, 187). As a consequence, RT can be classified as a non-mainstream or alternative news outlet (see also Boberg et al. 2020).

We expect the debates on the Facebook sites of both media outlets to differ widely since the nature of media reports on both sites is likely to vary, and since the readership of SZ and RT is likely to diverge significantly. The quality journalism of the SZ is likely to attract a more educated, centrist readership with little affinity to Covid-19-related conspiracy theories. The critical, non-mainstream journalism of RT, in contrast, is likely to find a less educated, politically more extreme readership which we expect to be more vulnerable to conspiracy theories.

In order to select suitable articles for analysis, we collected articles about the pandemic over a period of one week, which were published on the two selected social media outlets SZ and RT. The early stages of the pandemic, during which we collected data in an arbitrary week, have been marked by a high level of media coverage and little tangible knowledge about Covid-19. Due to the political and scientific relevance of the pandemic, we concentrated exclusively on articles dealing with the consequences of the pandemic and government measures adopted to combat the spreading of the virus. In addition, articles drawing too few user comments (less than 150 comments) were excluded from the selection process. This yielded three RT and four SZ articles. In order to have a comparable sample for both platforms, we randomly selected three out of the four SZ articles for analysis. To facilitate the preservation and processing of the comments, we extracted the comments of the respective articles into an offline document, resulting in a total of 1,104 collected comments (see Table A1 in the appendix for an overview of the articles and the number of retrieved comments). To protect the identity of the debaters, we anonymized the usernames prior to the analysis.

We are aware that the time frame of our analysis is relatively short, which does not allow us to say anything about developments over time. However, the aim of our analysis is not to trace how a potential process of polarization has unfolded over time. Instead, we deliberately take a snapshot view of the debates about Covid-19 during the onset of the pandemic in the spring of 2020. Should we find major traces of affective polarization in our material, we can assume that a process of polarization has taken place before, which produced the situation of polarization we observe.

To analyse the debates that developed around the articles on Covid-19, we used tools of qualitative discourse analysis. In our view, qualitative discourse analysis is highly suitable for the purpose of this study, as it is an appropriate method to obtain an in-depth understanding of the construction of meaning and the creation of collectively shared understandings of the world in public debates (Willig 2014; Angermuller 2015). As far as the analysis process is concerned, we started with a coding scheme that operationalizes our two layers of affective polarization: the emergence of hostile outgroup identities and the development of political intolerance towards the other camp (see Table A2 in the appendix for details of our coding scheme). Based on this coding scheme, we classified the Facebook comments of the selected articles within different discourse strands. These strands were then examined for recurring patterns and frequently used motifs. Coding and analysis were done with the help of the qualitative data analysis software Atlas.ti.

In order to assess the reliability of the content analysis, we conducted an intercoder reliability test. A reliability sample of ten per cent of data was randomly selected from the data. To measure the intercoder reliability of our content analysis, a widely-used and easy-to-calculate indicator was chosen: Holsti’s coefficient (Holsti 1969). The first step of content analysis resulted in a coefficient of 0.83. In a second phase, both coders discussed the codes on which they had not agreed. In all cases, the two coders came to a mutual agreement. These outcomes of the intercoder reliability test led us to the assumption that the content analysis of this study is reliable and not only the result of a subjective interpretation.

Empirical results

Our analysis of user comments on the Facebook platforms of RT and SZ showed clear signs of the development of a hostile outgroup identity, the first dimension of affective polarization. Our discourse analysis revealed that almost all people who participated in the online discourse on the RT platform shared a common group ideology. We detected very few substantive disagreements. Rather than contradict each other, the commenters reinforced each other in most discussions. Moreover, the debates on the RT platform were strongly characterized by stigmatization and the exclusion of dissenters. The vast majority of comments were marked by anger or rage at the actions of the other camp. Whenever the prevailing opinion was challenged, the supporters of this view did their utmost to impose their own worldview on the ‘renegades’. Common catchphrases such as ‘wake up!’ and calls to ‘read between the lines’, combined with web links to videos and articles revealing ‘the truth’, were used to set dissenters on the right path.

In this context, we also found clear indications of conspiracy theories, as shown, for example, in the following user comment: ‘He [Federal Health Minister Jens Spahn] has the Bilderbergers and Bill Gates behind him, they are good allies against the people’ (RT article 1). Our analysis shows that conspiracy thinking was often associated with extreme statements and affirmative responses by other users. In general, the vast majority of users on the RT platform believed that politicians are generally evil and do not serve the interests of citizens, as shown, for example, in the following user comment: ‘The judges of the Federal Constitutional Court are from their ranks. Forget it. Politics, judges, media, they are all the same’ (RT article 1).

The discourse on the SZ Facebook page also shows traces of the development of a common group identity, albeit in a different form than on the RT page. The SZ discourse was primarily dominated by a science-based ideology, although this view was regularly challenged by users with dissenting opinions. However, such dissenting opinions were usually countered with comments based on ‘scientific facts’.

In contrast to the discourse on the RT platform, where users often used short catchphrases, the debates on the SZ page proceeded in predominantly well-formed sentences with an argumentative structure juxtaposing different opinions on specific topics. The arguments put forward by commenters were generally reflective, aiming to contribute to better mutual understanding and constructive dialogue. An example of this is the way some SZ users criticized the Covid-19 measures taken by the government. In line with the more rationalistic discourse on the SZ, users proposed concrete measures that should be taken by politicians and substantiated this with arguments and statistics. It is interesting to note that the critical statements were usually not aimed at denigrating the responsible politicians, but at suggesting better and more effective ways of acting. This is in strong contrast to the discourse on the RT platform, where the debate centred around the idea that politicians are generally evil and do not want to serve the interests of citizens.

But the online discourse on the SZ platform also exhibited negative features of stigmatization and denigration, which were almost exclusively associated with a strong self-differentiation of mainstream SZ users from conspiracy theorists and Covid-19 deniers. In contrast to RT users, only a minority of SZ users expressed actual conspiracy theory beliefs in their comments, and if they did, the references were rather indirect. For example, by defending ‘critical thinking’ and ‘questioning’ the mainstream press, this minority of SZ users employed rhetorical elements that can also be found in conspiracy narratives. Such statements were usually met with fierce criticism from the mainstream SZ commenters, who immediately associated these arguments with conspiracy thinking and quickly isolated the dissenting minority from the dominant rationalist discourse. This becomes particularly clear when analysing the user comments to SZ article 2 entitled ‘Why so many people believe in Covid conspiracy myths’. Many SZ commenters displayed feelings of superiority when laying out their ideas about why people allegedly believe in conspiracy theories. For example, one user wrote: ‘In an increasingly complex world, people are looking for simple answers. Unfortunately, the real connections, explanations and solutions are not simple. That is why alternative facts come into play’ (SZ article 2).

In conjunction with the overall reflexive attitude of SZ users and the generally respectful tone dominating the SZ discourse, only a few ideologically extreme statements could be found in the user comments on the SZ platform. However, when ideologically extreme statements were voiced, they were highly controversial and usually provoked several counter-comments. These reactions included follow-up questions, efforts at appeasing the situation and attempts to refute the argument, but also in some cases insults and expressions of disrespect towards the author of the comment. Interestingly, SZ users directed derogatory behaviour almost exclusively toward committed Covid-19 deniers and conspiracy theorists.

Next to the formation of hostile outgroup identities, our data also showed the development of political intolerance towards the other camp on both platforms, the second dimension of affective polarization. On the RT platform, users frequently displayed extremely negative views of the other side, which went beyond mere rejection but instead suggested the presence of strong hostile feelings, with users often using aggressive and vitriolic expressions.

In extreme cases, the strong feelings of hostility towards the other camp turned into insults and threats. Such hateful remarks were not uncommon in the RT discourse, usually lashing out at ‘the elite’ or government officials, as can be seen for example in the following user comment: ‘If he [Health Minister Jens Spahn] carries on like this, he will soon be standing face to face with a murderer’ (RT article 1). We also found that debates quickly moved into emotional territory where disagreements easily gave rise to insults, both against dissenters and against the political elite. Moreover, commenters regularly discredited their opponents by using jokes and stereotypes, often with racist connotations.

Many RT commenters also suggested that the government was involved in sinister machinations, which the general public failed to notice. For example, users claimed that politicians colluded with ‘the evil world elites’, that these ‘evil elites’ strategically used the pandemic to increase their power, or that Covid-19 did not actually exist. By invoking constitutional principles such as the freedom of assembly or the right to bodily integrity against government measures such as the (possible) adoption of a vaccination mandate, some RT users accused politicians of breaking the law when implementing Covid-19 measures. Such claims were often accompanied by a strong sense of ‘us vs. them’ and even genuine anxiety about government action, as the following quote shows: ‘You get the feeling they really want to see dead Germans, don’t you?’ (RT article 1). While some RT commenters doubted that elections in Germany are fair, others even questioned the very existence of free elections.

Based on the conviction of sinister government machinations, some RT commenters called for taking action outside of the online world: ‘We are ruled by criminals! – When injustice becomes the law, resistance becomes a duty!!!’ (RT article 3). Examples like these clearly show that commenters on the RT site used moral justifications for their political claims, arguing that the moral evilness of their opponents left them with only one choice: violent protest and resistance. The most extreme calls to protest and resistance were directed against government measures. They are divided into three subgroups, which form an escalating ladder of increasingly radical claims: In the first stage, commenters called on their fellow debaters to stop acquiescing to supposedly repressive government measures. In the second stage, commenters called for moderate acts of political resistance, such as participating in demonstrations or boycotting the established parties. In the third stage, finally, commenters made the case for radical revolutionary action, often combined with concrete threats of violence against political elites. It also is remarkable that the data contain very few statements contradicting such claims or rejecting the predominantly negative view of politics on which they are based. The few users who spoke out against such claims immediately became the target of fierce criticism from the supporters of the prevailing view, often accompanied by personal attacks.

On the SZ platform, the online discourse also shows the development of a hostile attitude toward the other camp. This went beyond the stigmatization of conspiracy theorists per se. Users tended to question the mental sanity of members of the other side and often denounced every opponent as a conspiracy theorist. Comments such as ‘While viruses can easily spread in weakened bodies, the same happens with conspiracy theories in their weak minds’ (SZ article 2) show a very condescending attitude.

Furthermore, SZ commenters used arrogance, jokes and provocations to belittle their opponents. As the style of argumentation on the SZ platform was much more differentiated and the opinions less one-sided than on the RT platform, the constant depreciation of ‘the others’ was cleverly disguised, but the discourse nevertheless revealed a highly hostile outgroup identity towards dissenters. SZ users frequently ridiculed conspiracy theories and defamed the people who believe in them, portraying them as stupid and making fun of them with provocative comments such as: ‘You’re not very smart, are you?’ (SZ article 3). This contemptuous behaviour was often combined with feelings of moral superiority over the people who believe in such ‘nonsense’. Statements such as ‘Of course, Bill Gates is to blame for everything’ (SZ article 2) are an example of this phenomenon.

The attitude shown in the SZ comments goes beyond simply stigmatizing the other camp. The frequent devaluation and ridicule of conspiracy theorists and Covid deniers as well as the denigration of the other side’s positions as incomprehensible, illogical and inferior imply that the SZ discourse failed to accept the arguments and positions of ‘the others’ as morally legitimate. Mainstream SZ commenters did not take the views of the other side seriously, let alone accept them as a legitimate part of the political debate. Due to the low number of dissenting users on the SZ platform, this could be observed more in the way mainstream commenters talked about opponents than in the way they directly talked to opponents. However, this is not necessarily a sign of the absence of polarization but rather indicates a strong group identity on this platform, which serves as a place for like-minded people to interact. As the reactions to the few cases where users expressed dissent show, polarization did take place between the vast majority of like-minded people and ‘the others’, who were mostly active elsewhere – for example on the RT platform. These features of the discourse are a clear sign that the second stage of affective polarization, the development of political intolerance towards the other camp, was present in the SZ discourse – just like in the debates on the RT platform.

Discussion and conclusion

The aim of this article was to investigate the presence of affective polarization in online debates surrounding the Covid-19 pandemic in Germany. In the first part of this article, we argued that one can distinguish two stages of affective polarization: (1) the emergence of opposing groups with hostile outgroup identities toward each other, and (2) a more radical form of affective polarization additionally characterized by political intolerance, in which identity-based divisions become so strong that groups no longer accept each other as legitimate participants in democratic exchange. We argued that it is especially this second stage of affective polarization that is problematic for democracy.

Our analysis of debates surrounding the Covid-19 pandemic on the Facebook platforms of a mainstream and a non-mainstream news outlet in Germany, SZ and RT, focused on how users dealt with Covid-19-related conspiracy theories. Conspiracy theories and polarization are mutually dependent. Belief in conspiracy theories is inherently polarizing because it leads not only to the discrediting of any legitimacy of opposing views but also to the denial of science-based truths. At the same time, a polarized debate with an us-against-them mentality drives discourse to extremes, creating the breeding ground for the proliferation of conspiracy theory content.

In our analysis, we found patterns of both dimensions of affective polarization on each of the analyzed platforms (see Table 2 for an overview of the findings). Yet, as one might expect for debates following articles from a non-mainstream media outlet, we found the highest level of affective polarization in the online discourse on the RT platform. The RT discourse was characterized by widespread conspiracy beliefs, with users voicing highly emotional negative feelings towards an ‘omnipotent’ system run by ‘evil’ government elites instrumentalizing the pandemic to increase their power. Almost all comments showed signs of Stage 1 affective polarization by developing a hostile identity towards the political opponent. Moreover, a large number of comments went beyond this and clearly showed political intolerance towards the other camp. This second stage of affective polarization manifested itself in the RT discourse in strong hostile feelings as well as insults and threats. Users were convinced that the government was involved in sinister machinations and called for radical or even revolutionary action. Many commenters on the RT platform espoused conspiracy theories about the spread of the Covid-19 virus and the supposed sinister motives behind the government's control measures. The posts were often highly emotional and led to heated debates between supporters and opponents of individual claims, peppered with insults and threats. At the heart of these debates were strong feelings of distrust toward a seemingly all-powerful ‘system’ run by ‘evil’ politicians. Users who slightly challenged these views defended the government’s actions or doubted the validity of conspiracy theories faced harsh, emotionally charged responses. Users who expressed such views were often treated condescendingly and stigmatized as ‘leftist government supporters’. This shows how an extreme oppositional stance to the norms of democratic society has become consensus. As a result, features of the second, more harmful stage of affective polarization, combining negative outgroup identities with political intolerance toward the other camp, were a key feature of the online discourse on RT.

Table 2. Affective polarization on the Facebook platforms of RT and SZ.

	Russia Today Germany (RT)	Süddeutsche Zeitung (SZ)
Hostile outgroup identities (Stage 1)	­ Discourse dominated by an anti-establishment ideology and conspiracy beliefs; few episodes of disagreement ­ Stigmatization and exclusion of dissenting opinions ­ Frequent ideologically extreme statements; general agreement with such statements by others	­ Discourse dominated by a science-based rationalist ideology; regular episodes of disagreement ­ Stigmatization and exclusion of dissenting opinions ­ Few ideologically extreme statements; if they were expressed, other users seriously objected
Political intolerance towards the other camp (Stage 2)	­ Extremely negative picture of the other side, with strong hostile feelings expressed through insults and threats ­ Conviction that the government is involved in sinister machinations, which the opposing side fails to notice ­ Calls for radical revolutionary action to counter the government’s conspiracies	­ Opposing side denounced as stupid believers in conspiracy theorists ­ Use of arrogance, ridicule and provocation vis-à-vis opponents ­ Failure to accept the positions of the opposing side as morally legitimate

In contrast, the SZ discourse was much calmer and more rational, virtually free of direct conspiracy views. In contrast to the RT discourse, the debates on the Facebook platform of SZ were far less confrontational. Consistent with the ideologically centrist, mainstream nature of SZ media coverage, users who debated on the newspaper’s Facebook platform expressed less extreme views than their RT counterparts. Debates were dominated by rational arguments about the spread of Covid-19 and what to do about it. Only very rarely did we find contributions characterized by (indirect and relatively mild forms of) conspiracy thinking, so the discourse was not marked by episodes of open confrontation between supporters and opponents of conspiracy theories.

Despite the generally moderate and rational character of the debates, however, our discourse analysis also revealed evidence of both stages of affective polarization in the SZ discourse, manifested in hostile references to supporters of conspiracy theories outside the SZ platform. SZ users denounced the views of conspiracy theorists as naive and unscientific, referring to them merely as ‘the others’ – an outgroup from which members of the rationalist camp should distance themselves. This distancing from ‘the others’ went hand in hand with the second dimension of affective polarization: like mainstream RT users, SZ users considered the views of their opponents to be completely illegitimate. Sceptics may argue that SZ users, unlike RT users, rightly rejected the claims of conspiracy theorists on the basis of prejudice or insufficient evidence and that rejecting the content of someone else's claims on rational grounds is part of a lively democratic discourse. However, this argument does not justify how SZ users reacted to dissenting opinions. Rather than pointing to the lack of rational reasoning and scientific evidence, mainstream SZ users immediately jumped to very negative conclusions, mocking the naive views of conspiracy theorists and making sarcastic jokes about them. SZ users thus rejected the opinions of dissenters as illegitimate without bothering to discuss to what extent these positions may fall outside of the scope of rational, science-based reasoning.

What is also interesting about these results is that the politically disaffected conspiracy theorists who dominated the discourse on RT bear a strong resemblance to ‘the others’ from whom the dominant SZ discourse tried so hard to distinguish itself. And the arguments of the defenders of political reason, who were in the minority in the RT debates, share many important features with the dominant discourse on the SZ platform. As a result, these two opposing discourses stabilized and reinforced each other, deepening the gap between them step by step. In this sense, both discourses undermine a cornerstone of even the most conflict-oriented notions of democracy: the recognition that opposing political positions have a legitimate place in public debates and that, even if not everyone has to share all these positions, they have a right to be expressed and discussed in the democratic process. Our findings thus show that open and fair democratic debates can be threatened not only by radical extremists and conspiracy theorists but also by the intolerance of more moderate rationalists in the middle of the political spectrum.

Finally, the contribution of our study to the wider literature on political polarization and democracy is twofold. First, our work contributes to the literature by expanding the conceptualization of affective polarization. We argue that further research might benefit from conceptualizing affective polarization as a two-stage concept: starting with emotional resentment rooted in identity politics, and evolving into a second, more radical, stage where identity-based hostility turns into political intolerance toward opponents. Stage 1 affective polarization seems like a rather normal phenomenon in democracies marked by a certain level of heterogeneity and, thus, political conflict. Even though it is incompatible with the consensus-based approach of deliberative democracy, it is partly or fully in line with more adversarial theories of democracy like pluralist democracy and, in particular, agonistic democracy. The fully-fledged version of Stage 2 affective polarization, in contrast, is problematic from the perspective of even the most conflict-oriented approach to democracy: agonistic democracy. Deep ideological divisions and deeply-held hostile outgroup identities leading to negative emotions toward outgroups could still be seen as signs of a functioning system of democratic competition. Alarm bells should, however, start ringing as soon as outgroup resentment turns into open political intolerance. When adversaries turn into outright enemies who do not accept each other as legitimate political players anymore, this poses a serious threat to peaceful democratic exchange.

Second, in contrast to recent studies on affective polarization, which have tended to focus heavily on the polarization of political attitudes using techniques from quantitative research (e.g. traditional survey-based methods as well as social network analysis; see, for example, Lelkes 2016; Lelkes, Sood, and Iyengar 2017; Suhay, Bello-Pardo, and Maurer 2018; Iyengar et al. 2019; Boxell, Gentzkow, and Shapiro 2020; Asimovic et al. 2021), we contribute to the smaller body of qualitative research on affective polarization at the level of online political behaviour by using qualitative online discourse analysis (see also Tucker et al. 2018, 11; for other qualitative work in this field, see Steuter and Wills 2009; Krochik and Jost 2011; Harel, Jameson, and Maoz 2020; Wodak and Rheindorf 2022). The richness of this qualitative approach allowed us to explore in-depth the role of outright political intolerance in people’s affective attitudes toward their opponents, gaining new insights into how affective polarization may unfold differently in online debates. This in-depth analysis would have been difficult to achieve in a quantitative study that could have covered more material but would have been limited to classifying texts into broader categories (see, for example, Yarchi et al. 2021). Yet, we believe that our focus on different stages of affective polarization, which have different implications for the viability of democratic exchange, is a fruitful contribution to both qualitative and quantitative scholarship on the topic in the future.

The downside of our methodological approach is that our findings are based on a rather small sample of online comments and are, therefore, hard to generalize. It might be that the crisis situation of the Covid-19 pandemic created a societal situation that was particularly conducive to patterns of affective polarization. It might also be that debates on other social networking sites had a different dynamic than the debates on Facebook, and it is well conceivable that our results from Germany may not easily travel to other countries’ contexts. More research on affective polarization in online debates is thus needed to explore how typical or atypical our findings are. However, a growing body of empirical evidence leads us to believe that our findings may be more typical of political debates than the small sample suggests. On the one hand, evidence from German media reports about anti-Covid and anti-vax demonstrations (see, for example, Callison and Slobodian 2021), and the reactions to such events by journalists, politicians, and talk show guests, indicate that the type of politically intolerant debate we uncover in this article is certainly more widespread in Germany. On the other hand, the fact that we find major indications of political intolerance in online debates in a country like Germany, which has been known for relatively low levels of affective polarization in the past (Boxell, Gentzkow, and Shapiro 2020) suggests that we could expect similar, or even much larger, levels of affective polarization in other countries. Again, empirical evidence from countries like the US or France suggests that very similar types of debates have been going on not only in Germany but also elsewhere (see, for example, Druckman et al. 2021).

Apart from studies analysing the polarizing nature of online debates in other contexts, we think that a promising strategy for future research would be to explore how and why (different stages of) affective polarization emerge in political debates. Are there specific social, psychological, or ideological conditions that drive people to engage in affectively polarizing debates? What role do political context conditions such as different forms of political communication by political parties and the media play? When and how do debates shift from emotional negativity to outright intolerance? And finally, once political intolerance has taken hold, are there ways of putting the genie back in the bottle?

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 According to OECD data, the confidence in the national government reached 65% in Germany in 2020, compared to an OECD average of around 50%. Similar to the trend in most other OECD countries, trust in the government in Germany declined during the first COVID-19 wave in 2020, yet at a somewhat lower rate as in many other OECD countries (OECD 2021).

2 See https://www.statista.com/statistics/1059426/social-media-usage-germany/ (accessed on 23 August 2022).

3 For the data on the overall reach of Süddeutsche Zeitung, see https://de.statista.com/statistik/daten/studie/381931/umfrage/reichweite-der-sueddeutschen-zeitung/. The number of Facebook subscribers refers to April 2020: https://web.archive.org/web/20200426173206/https://www.facebook.com/ihre.sz/ (both accessed on 23 August 2022).

4 See https://www.mediadb.eu/forum/zeitungsportraets/sueddeutsche-zeitung.html (accessed on 23 August 2022).

5 See https://deutsch.rt.com/uber-uns/ (accessed on 23 August 2022).