Josef Šlerka and Vít Šisler, Charles: Who is Shaping Your Agenda? Social network analysis of anti-Islam and anti-immigration movement audiences on Czech Facebook
This study was originaly published in: Expressions of Radicalization: Global Politics, Processes and Practices, edited by Kristian Steiner and Andreas Onnerfors, Palgrave Macmillan, 2018, pp. 61-85.
The internet has become a crucial tool for people with similar interests to reach out to each other for information and support, to share ideas, and to create personal networks (Rainie and Wellman, 2012:107). Several researchers have suggested that online interactions and materials should be considered key elements in radicalization (Bouchard and Levey, 2015:2). Indeed, radical movements increasingly use the internet to advance their goals (Bouchard and Nash, 2015:53). Beyond easy access, little or no regulation and censorship, the anonymity of communication, and the fast flow of information, the internet offers the ability to shape coverage in traditional mass media (Weimann, 2006:30).
The democratization of mobile internet access around the world and the emergence of Web 2.0 have led to a more user-centric online environment. Radical movements can now rely on a crowd of anonymous sympathizers who are collectively engaged in the virtual dissemination of narratives and media content supporting their cause (Ducol, 2015: 84). Social media extend the traditional frontiers of radical online milieus by blending into platforms such as YouTube, Twitter, and Facebook (see Weimann, 2010). Today, any understanding of radicalization must take into consideration the impact (or lack thereof) of social media on social settings, media consumption, and the production of knowledge.
As Steiner and Önnerfors (2017) write in the introduction to this volume, radicalization is a multi-faceted, dynamic, processual, and multidimensional phenomenon that defies easy definition. By the same token, Schmid (2013:7) notes that “radical” is a relative concept whose meaning has changed over time. As Neumann (2013:876) argues, radicalization has no obvious essential or inherent characteristics; rather, it is a process of positioning relative to chosen points of reference. As such, it can only be understood in terms of its distance from these points of reference, be they “status quo or mainstream positions on the political spectrum of a given society” (Schmid, 2013:56).
In this chapter, we study radicalization as a social process and communicative practice, where “radical ideas are transmitted by social networks” (Dalgaard-Nielsen, 2010:803) and contribute to the polarization of public political discourse. Such radicalization processes include, yet are not limited to, advocating sweeping political change and system-transforming solutions for government and society that depart from “the democratic rule of law and international human rights standards” (Schmid, 2013:8). This chapter focuses on the radicalization of public discourse in light of the contemporary migration crisis and on the role of anti-Islamic and anti-immigration movements in shaping the agendas of mainstream media through social network sites.
User-generated data connected to the proliferation of social media are growing exponentially. Analysis of these “big social data” opens up new perspectives for research in social sciences and the humanities (boyd and Crawford, 2012; Halavais, 2015; Manovich, 2011). Meanwhile, advances in information and computer technologies present new research methods and new approaches. Being able to quantitatively process large datasets through automation opens a path to new research questions and new ways to answer them (Šlerka and Šisler, 2017).
Manovich (2011) argues that the rise of social media and advances in computing tools to process massive amounts of data make possible a fundamentally new approach to the study of human beings and society. We no longer must choose between data size and data depth. We can study exact trajectories formed by billions of cultural expressions, experiences, texts, and links. The detailed knowledge and insights that before could only be gathered about a few can now be gathered about many. In Manovich’s (2011) terms, we are no longer forced to choose between surface data about the many or deep data about a few.
In light of current debate about the migration crisis and the simultaneous proliferation of radical movements in the public sphere, we are in crucial need of critical investigation of the structural and dynamic aspects of audience formation and agenda shaping on social network sites. With Bouchard and Levey (2015:4), we believe that integrating network concepts and network methods into the study of radicalization is fundamentally important (from theoretical, empirical, and policy perspectives) to bringing the field forward. The network methods approach allows the structural aspects of various groups to be accurately depicted without potentially false assumptions about the ways these groups function. It lets patterns and unexpected findings to emerge from the data (Bouchard and Nash, 2015:50).
Digital media can “support the formation of a public sphere, where a diversity of opinion and information can interact, or, conversely, to function as an echo chamber that reinforces established perspectives and opinions” (Colleoni, Rozza and Arvidsson, 2014:317). Given that both these scenarios are well established and simultaneously contested in the research on political communication on the internet (see e.g. Brundidge, 2010; Stroud, 2010), exploratory research analysing big social data through network methods is particularly viable for enhancing our understanding of online radicalization. It is also very important to ensure that our research methods are transparent to allow other researchers to engage with, replicate, and possibly falsify our research and findings.
This chapter presents exploratory research on the social network sites of Czech anti-immigration and anti-Islamic movements. It analyses the audiences of these movements’ sites on Facebook and explores their similarities, differences, and affinities through social distance computed based on their fans’ likes. The chapter uses the new, formally defined, quantitative method Normalized Social Distance (NSD) developed by Šlerka (2013) and detailed by Šlerka and Šisler (2017). NSD calculates the distance between various social groups based on members’ intentional stances as expressed on social networking sites. NSD provides an opportunity for distant reading of social network sites, enabling us to formally represent and analyse the structural aspects of big social data.
The primary aim of this chapter is to investigate how near to, or far from, each other Facebook audiences of Czech anti-immigration and anti-Islam movements are in terms of NSD. We also analyse their distance from major Czech news media, established political parties, and politicians active in public debate about the migration crisis. The secondary aim of this chapter is to examine the structural interplays between Czech anti-immigration and anti-Islamic movements’ active audiences and the news media’s shaping of agendas on Facebook. Through quantitative analysis of the most popular posts, we explore how diverse audiences elevate particular stories on Czech news media through the distribution of likes on Facebook.
More generally, this chapter aims to present a new methodological framework for the analysis of big social data, especially data from Facebook. The case study itself serves as an example of the method using a concrete dataset, which explains or clarifies possible further interpretative approaches. The methods proposed in this chapter constitute a coherent set of tools, which could be adopted relatively easily by a variety of actors to support their research or decisions with empirical evidence.
Social Network Sites and Self-Representative Performance
The term “social media” conveys several meanings. Most authors in the field agree that social media constitute a virtual space in which the possibility of social interaction between users plays a crucial role and these interactions have a specific impact on the creation of user identities, communication situations, and communities (boyd and Ellison, 2007; Obar and Wildman, 2015). Nevertheless, the term blurs the distinction between different platforms and communication channels (Obar and Wildman, 2015:746).
Given the analytical ambiguity of “social media” we have opted to use the term “social networking sites” instead. In accordance with boyd and Ellison (2007), we define social networking sites as web-based services that allow individuals to (1) construct a public or semi-public profile within a defined system, (2) articulate a list of other users with whom they share a connection, and (3) view and go through their list of connections and those made by others within the system.
We have argued elsewhere that these connections shared with other users on social networking sites are the result of social actions and possess an intentionality of their own (Šlerka and Šisler, 2017). The user’s behaviour in social networks is not only a social action taken towards others, but also a representation of an intentionality that presupposes other subjects and anticipates their interpretations of such behaviour. The analysis of actions on social networking sites is thus an analysis of data representing not only certain behaviours, but also “intentional stances” (Dennett, 1996).
If we understand user behaviour on social networking sites within Goffman’s (1959) framework of dramaturgical sociology, the user’s profile and social action conducted through that profile could be considered part of a “personal facade” or as actions happening on the “front stage.” From this perspective, all the elements that form a personal profile on a social network site are elements of the facade that users select to represent their personal identity. The choice of name, profile photo, description, privacy settings, etc. could all be perceived as expressions of the user’s identity and front stage performance.
Social actions conducted on social networking sites (e.g., status posts, comments, “likes” of other users’ pages or posts) are forms of self-representative performance (Wallace, Buil and de Chernatony, 2012). Social networking sites can thus be seen both as spaces for daily self-presentation and stages for performance and interaction. Actions on social networking sites have intentionality and can be analysed both quantitatively and qualitatively. From the perspectives of digital humanities and automated computational processing, it is quite possible to process all these social actions in an exploratory manner and search for structural patterns in the resulting data (Šlerka and Šisler, 2017).
Existing research on social networking sites relevant for our study can be divided roughly into three research clusters: (1) users’ online behaviour, (2) media consumption and agenda shaping, and (3) online radicalization.
Kosinski, Stillwell, and Graepel (2013) demonstrate how publicly accessible information about users’ Facebook likes can be used to predict automatically and accurately a range of highly sensitive personal attributes including sexual orientation, ethnicity, religious and political views, personality traits, intelligence, happiness, use of addictive substances, parental separation, age, and gender. Taking a different tack, Pelletier and Horky (2013) present exploratory qualitative research to look at the motivations and consequences associated with liking commercial brands’ pages on Facebook. Wallace et al. (2014) similarly explore a typology of fans (i.e., individuals who like different brands’ pages on Facebook). Recent exploratory qualitative studies have investigated individual users’ motivations to like their friends’ posts on Facebook (Basalingappa, Subhas and Tapariya, 2015).
Social networking sites may play a significant role in how people gather political information (Bode, 2016). Social ties play a major role how the public learns about politics, offline social networks play a role in the dissemination of information (Ellison and Fudenberg, 1995), and information from trusted people is deemed more credible and is more likely to be taken seriously (Huckfeldt, Beck, Dalton, and Levine, 1995:1027). Today, exposure to political information within social networking sites is much like that from the sources that came before them, such as news websites and more traditional media (Bode, 2012). However, research suggests that the potential for users to gather political information from social media is not always realized within the general population (Bode, 2016). News publishers take social networking sites seriously and include them in their media strategy. A recent report published by Parse.ly (2015), an analytics firm that collects data for digital publishers, suggests that Facebook already drives more traffic to news media websites than Google. Consequently, larger news and media sites have become much more reliant on Facebook and shape their editorial policies accordingly (Ingram, 2015).
Content on Facebook’s News Feed is selected by algorithms based on a user’s previous behaviours (Pariser, 2011) and individuals are increasingly exposed to information from like-minded individuals (Flaxman, Goel and Rao, 2016), leading to renewed speculation about “echo chambers” and “filter bubbles” devoid of attitude-challenging content (Bakshy, Messing and Adamic, 2015:1130). Increasing reliance on Facebook as a gateway to news media could lead to reaffirmation of people’s existing political orientations. In their seminal study, Bakshy et al. (2015) examined how 10.1 million US Facebook users interacted with socially shared news. They measured ideological homophily in friend networks and the extent to which heterogeneous friends could expose others to cross-cutting content. Their findings suggest that (1) with Facebook’s automatic ranking of posts, people on average have slightly less cross-cutting content in their News Feed, and (2) exposure to ideologically different content is further limited by individual choices (Bakshy et al., 2015:1131). Nevertheless, “despite these tendencies, there is substantial room for individuals to consume more media from the other side; on average, viewers clicked on 7% of hard content available in their feeds” (Bakshy et al., 2015:1131). In other words, rather than people browsing only ideologically aligned news sources or opting out of hard news altogether, Bakshy et al.’s research shows that social network sites “expose individuals to at least some ideologically cross-cutting viewpoints” (2015:1132).
A limited, albeit growing, body of research on social network sites addresses online radicalization. Social ties and social influence have been found to be central to the radicalization process (Hegghammer, 2006; Sageman, 2004, 2008). Social network sites are used by various radical movements to spread beliefs and ideologies, recruit members, and create online virtual communities with a common agenda (Agarwal, 2015). As Ducol (2015:86) argues, interactive features of modern web-based technologies, including social networking sites, have facilitated a broader dissemination of autonomous, user-generated content outside official websites and digital platforms. Meanwhile, they have also eased the emergence of undefined online communities, radical digital milieus (Conway, 2012), that encompass a broad cross-section of producers and consumers who all contribute to the everyday re-making and dissemination of radical narratives through cyberspace.
Agarwal’s (2015) comprehensive review of research on online radicalization includes characterization, classification, and an in-depth meta-analysis of about 100 conference and journal papers published over the past 10 years, revealing that most such studies target events specific to a country or region (mainly USA and Latin America) and mine English language texts. Most studies use a variety of information retrieval methods, automated text processing, and methods of analysis based on machine learning.
As far as we know, no other study has used NSD to analyse social networking sites’ audiences to examine online radicalization and none has discussed anti-Islamic and anti-immigration movements on Facebook in the Czech Republic.
Normalized Social Distance
The concept of NSD was introduced by Šlerka (2013) and detailed by Šlerka and Šisler (2017). For the sake of brevity, we describe only key features of NSD here and refer readers to the above-mentioned studies for details.
NSD is a formally defined method that calculates distance between social groups based on intentional stances expressed in group members’ activities on social networking sites – in our case, on Facebook pages. The resulting number expresses how far or close various sites’ audiences are in relation to each another. Importantly, NSD relies on post likes (i.e., likes given to specific posts published by the page in question) rather than page likes (i.e., likes given to a page in general). This methodological distinction assumes that while a page like could represent a variety of intentional stances ranging from interest in the page’s activity to support of the ideas expressed, a post like probably expresses affirmation of the ideas in a particular post (see Wallace et al., 2014).
Theoretically, NSD stems from McPherson, Smith-Lovin, and Cook’s (2001) concept of homophily in social networks, from Lin’s (1998) information-theoretic definition of similarity, and particularly from Cilibrasi and Vitányi’s (2010) concept of normalized web distance.
Homophily is the principle that contact is more frequent between similar people than dissimilar people (McPherson et al., 2001); it assumes that similarity breeds connection. The homophily principle structures network ties of every type including marriage, friendship, work, advice, support, information transfer, co-membership, etc. The result is that people’s personal networks tend to be homogeneous in many sociodemographic, behavioural, and intrapersonal characteristics. Homophily limits people’s social worlds in a way that has powerful implications for the information they receive, the attitudes they form, and the interactions they experience. Homophily also implies that distance in social characteristics translates into network distance, the number of relationships through which a piece of information must travel to connect two individuals (McPherson et al., 2001).
Introduced by Cilibrasi and Vitányi (2010), normalized web distance (NWD) is a semantic measure of similarity derived from the number of hits returned by an internet search engine for a given set of keywords. Words or phrases with the same or similar meanings (in a natural language sense) tend to be close in units of web distance, while words with dissimilar meanings tend to be further apart. We can perceive NWD as an expression of semantic distance. Using internet search engines, particularly Google, NWD often relies on contexts expressing a large body of common-sense knowledge. In a series of experiments, the accuracy of NWD was evaluated against expert opinion with positive results (Cilibrasi and Vitányi, 2010).
NWD comes with the idea of a semantic layer of information, but we propose that there is another, more pragmatic, layer above it that depends on the degree of similarity expressed by the online behaviour of two different social groups. We can also formally define the method of calculating the distance between these two social groups (Šlerka and Šisler, 2017).
Formally, we define NSD as follows, where f(x) is the number of members in the group x; f(y) is the number of members in the group y; f(x, y) is the number of elements that are simultaneously members of both groups; and N is the number of all elements in the given corpus:
With distance thus formally defined, we should be able to measure the distance between any two social subgroups that fall under the umbrella of another one. NSD is a universal metric that can be adjusted for data from any social network sites. In the following case study, we applied the formula to Facebook pages; calculating the proximity of these pages based on post likes from these pages’ active users. The resulting matrix is a bimodal network with a relatively low density that can be examined using traditional exploratory techniques such as hierarchical cluster analysis, multidimensional scaling, or principal component analysis (Šlerka and Šisler, 2017).
Clustering of Anti-Immigration and Anti-Islam Movements on Czech Facebook
The first aim of this study was to conduct exploratory research on anti-Islamic and anti-immigration movements on Czech Facebook. Primarily, we analyse how close or far audiences of these movements are in relation to each other in terms of NSD. Secondarily, we analyse the proximity of these movements’ audiences to key Czech news media, established political parties, and politicians active in the public debate on the migration crisis.
We identified 56 Facebook pages for Czech anti-immigration and anti-Islamic movements, news media, political parties and movements, think tanks, campaigns, and individual politicians who are active in the public debate on the migration crisis. A list of these pages is provided in the Appendix.
For clarity, we have translated these pages’ names into English wherever possible (e.g. Green Party) or labelled these pages according to the following key:
(n) = news media
(m) = political movement or party
(p) = individual politician
In specific cases, we provide a full description of the page in parentheses (e.g., Miloš Zeman [President]).
We adopted the following procedure to compute the NSD of the selected pages to each another:
We downloaded all public posts by all the pages’ administrators between 1 September 2015 and 28 December 2015 for a total of 19,321 posts.
We downloaded a complete list of 540,775 unique online identifications (IDs) for those liking at least one of these posts. These users distributed 3,351,034 likes among the 19,321 posts.
Based on information from Facebook Audience Insights (2016), we estimated the Czech Facebook region to include 3,500,000 unique users.
Based on these data, we computed the NSD between all the selected Facebook pages.
The results of NSD can be visualized in several ways. For this study, we combined graphic visualization and clustering analysis.
Figure 1 depicts only significant pages (i.e., those with significantly overlapping audiences in terms of NSD). The nodes in the graph denote individual pages; the links denote significant proximity in terms of NSD (i.e., NSD(x,y) ≤0.5).
Figure 1. Anti-immigration and anti-Islamic movements’ proximity to media, politicians, and political parties on Czech Facebook based on the NSD metric.
Figure 2 illustrates our clustering (k-means) clustering analysis. This method aims to partition our observations into clusters in which each observation belongs to the cluster with the nearest mean, which serves as a prototype for the cluster. This results from partitioning the data space into Voronoi cells. (The results of NSD analysis are multidimensional, and their visualization in a two-dimensional space should be understood as a mathematical approximation.)
Figure 2. k-means clustering of anti-Islamic and anti-immigration movements, news media, politicians, and political parties on Czech Facebook based on the NSD metric.
NSD analysis provides us with an opportunity for the distant reading of social network sites and their audiences. This distant reading clarifies structural aspects not necessarily visible on the level of “lose reading, (i.e., content analysis or interviews).
In our case study, the results of NSD analysis (in both the graph visualization and the clustering analysis) show several key findings about the anti-Islamic and anti-immigration movements and the proximity of these pages’ audiences to one another; as well as to Czech news media sites and the sites of Czech political parties.
The findings show several tightly connected clusters of pages on Czech Facebook whose audiences are significantly close to one another and share similar intentional stances. The users in these clusters like and share similar content and rarely reach out to different clusters.
Among these clusters, we can identify one, which could be labelled as anti-Islamic, anti-immigrant, nationalist, and/or anti-European Union (EU). This cluster consists of audiences active on the pages of the anti-Islamic political movements, Block against Islam and Stop Islam in the Czech Republic; the anti-EU and nationalist party, Freedom and Direct Democracy; and the libertarian/conservative Free Citizens’ Party. All these parties use strong anti-immigration rhetoric. Importantly, this cluster includes Miloš Zeman, President of the Czech Republic, who is known for his strong anti-Islamic and anti-immigration discourse. The news site Parlamentní listy, which also plays a prominent role in this cluster, publishes the un-redacted opinions of politicians and authors from across the political spectrum, yet still commonly linked to conservative and nationalist media. These pages’ audiences are very close to each other in NSD terms and show a significant overlap. They tend to rely on similar or close news sources and to like significantly similar content.
Another cluster consists of the active audiences of the liberal/left Green Party, the liberal Pirate Party, and an anti-discrimination campaign, Hate Free Culture. These pages’ audiences are close to the liberal weekly Respekt and the liberal daily newspaper Hospodářské noviny. As in the previous cluster, these audiences tend to rely on similar or close news sources and to like significantly similar content on Facebook.
At first glance, the public debate on the migration crisis seems highly polarized in the Czech Republic. News media tend to portray Czech society as fundamentally divided into two camps, corresponding roughly to the two audience clusters mentioned above on Czech Facebook (Šlerka, 2016). Similarly, many politicians perceive the migration crisis to be a divisive topic that can score them significant political points and they use it as such.
The NSD analysis reveals that, although these two audience clusters are significant and very active on other pages on Facebook, there are at least two other similarly significant clusters that seem to be primarily unrelated to the migration crisis debate and are rarely mentioned in the news media. These two clusters include, first, the audience of the ANO (“YES”) political movement and its leader, Andrej Babiš (also the Czech finance minister). ANO is a relatively new political entity that has often been portrayed as anti-establishment or populist; positioning itself as an alternative to the older “corrupt” parties. It has a vaguely defined programme and primarily promotes the “proper” technocratic management of public affairs (Šlerka, 2016). The other significant group unrelated to the migration debate is the audience of the Czech Social Democratic Party and Prime Minister Bohuslav Sobotka. Importantly, these two parties currently make up the government and lead the polls. These clusters have no significant proximity to any particular media outlets and have no significant overlap with the two active clusters identified earlier.
In summary, the NSD analysis reveals that, although public debate on the immigration crisis seems highly polarized into two adversarial clusters, it is more significantly fragmented in at least four different clusters, whose audiences rarely share the same content and whose intentional stances, as manifested by Facebook likes, rarely overlap. This structural fragmentation negatively influences public debate, while, in Habermasian (1989) terms, the possibility of communicative actions and mutual reasoning is seriously limited.
NSD is a quantitative method best suited to exploratory research. Unlike semantic methods in the digital humanities, NSD is featureless and is in principle unrelated to the content of the data analysed. It focuses on the actions of social network sites’ audiences (typically Facebook likes) and computes the distance between the audiences of different sites based on these actions. The assumption of the NSD method is that user behaviours on social networks are not only social actions, but representations of intentionality that presuppose other subjects and anticipate their interpretation of such behaviours. The analysis of actions on social network sites is thus the analysis of data representing not only certain behaviours, but also the intentional stances they represent. The NSD method allows falsification of results through qualitative analysis of the content users share and like. The falsifications the authors of this chapter have conducted so far suggest a possibly significant correlation between NSD and qualitative analysis, but further research is needed to confirm or refute this.
Post Overlaps Related to the Migration Crisis Debate on Czech Facebook
The second aim of the study was to analyse the structural interplays between the active audiences of anti-Islamic and anti-immigration movements and the agenda-shaping of Czech news media on Facebook. Through quantitative analysis of the most popular posts, this part of the study explored how diverse audiences elevate particular news on Czech news media Facebook pages through likes.
We used the same dataset of Facebook pages as in the previous section (i.e., the 56 Facebook pages of Czech anti-immigration and anti-Islamic movements, news media, political parties and movements, think tanks, campaigns, and individual politicians active in the public debate on the migration crisis listed in the Appendix).
In the second part of the study, we used but a more straightforward, quantitative analysis of post overlaps than the NSD:
We downloaded all the posts by the pages’ administrators from September and October 2015 and all unique user IDs for those who liked at least one of these posts.
We filtered these posts based on two additional criteria: popularity and overlaps. For the final data sets we selected only posts that (a) gained at least 50 likes in each selected month and (b) had at least a 15% overlap in likes with at least one other page on the list.
We treated the data from the two months as separate datasets to compare their structural patterns. The final dataset from September 2015 consists of 6,554 posts that attracted 1,072,425 likes from 261,833 unique users. The final dataset from October 2015 consists of 6,918 posts that attracted 930,570 likes from 220,575 unique users.
We computed the percentage of likes for individual posts from users who simultaneously liked another post on a different page in the dataset during the given period. From this basic matrix, we computed percentage overlaps among all the pages in the dataset. The algorithm for the computation is detailed in Šlerka (2016).
The results of the post overlap analysis are twofold: First, we can visualize the complete data as a correlation matrix, using the Pearson product-moment correlation. The Pearson correlation is a measure of the linear correlation between two variables x and y, giving a value between +1 and −1 inclusive, where +1 is a total positive correlation, 0 is no correlation, and −1 is a total negative correlation. The Pearson correlation is a measure of the degree of linear dependence between two variables. We have visualized the significant post overlaps among the pages in the dataset from September 2015 (Figure 3) and October 2015 (Figure 4).
Figure 3. Post/page overlap in September 2015.
Figure 4. Post/page overlap in October 2015.
Second, we can visualize the results as a table, listing all the posts with significant. For brevity, we include only one detailed example here. The complete results can be found in Šlerka (2016).
In the following example, we analysed Facebook posts from ČT24, a Czech national TV station operating as a public broadcasting service, and their overlaps with other audiences active in the migration crisis debate. In October 2015, ČT24 posted 1,301 posts on its Facebook page and an average post gained 169.5 likes (median = 64.0). From these posts, only 116 posts gained at least 300 likes and had at least 20% overlap with the audiences of other pages. These could be labelled as “trending” posts that were highly visible on Czech Facebook. Most (93) of these post overlaps were between ČT24 and other news media, and we therefore excluded them from our analysis. From the remaining 23 posts, 14 were related to the migration crisis debate and had significant overlaps with other active audiences (Table 1).
Table 1. Post/page overlaps on ČT24
Posting page Page with overlap Name of the post Total likes Likes overlap (%) ČT24 (n) Stop Islam in the Czech Republic (m) Eurosceptic party, Alternative for Germany (AFD), wants to file a criminal complaint against Chancellor Angela Merkel. 387 37.47 ČT24 (n) Stop Islam in the Czech Republic (m) Due to the open policy towards refugees, the German Public Prosecutor received hundreds of criminal complaints against Merkel. 395 36.96 ČT24 (n) Stop Islam in the Czech Republic (m) The Bulgarian border police shot dead a refugee, who illegally crossed the border, on the border with Turkey. 360 33,06 ČT24 (n) Stop Islam in the Czech Republic (m) Three-fifths of Czechs aren’t satisfied with EU membership. 330 32.73 ČT24 (n) Stop Islam in the Czech Republic (m) Harsh criticism for Merkel during a CDU regional conference. 314 31.53 ČT24 (n) Stop Islam in the Czech Republic (m) The Czech Republic will send two dozen soldiers and special military equipment to help the Hungarian army. 447 26.17 ČT24 (n) Stop Islam in the Czech Republic (m) Czech Army sends 650 soldiers to the Austrian border. 340 25.00 ČT24 (n) TOP 09 (m) President Zeman scares the public with his statements about refugees. 574 24.39 ČT24 (n) Hate Free Culture Eastern Europe should show more solidarity with refugees. 453 24.28 ČT24 (n) TOP 09 (m) Pope Francis denounces vicious campaigns waged against refugees in Europe. 508 21.65 ČT24 (n) Tomio Okamura (p) Three-fifths of Czechs aren’t satisfied with EU membership. 330 21.52 ČT24 (n) Tomio Okamura (p) Due to the open policy towards refugees, the German Public Prosecutor received hundreds of criminal complaints against Merkel. 395 21.27 ČT24 (n) Hate Free Culture Pope Francis denounces vicious campaigns waged against refugees in Europe. 508 20.67 ČT24 (n) TOP 09 (m) Eastern Europe should show more solidarity with refugees. 453 20.09
The findings in the second part of the study are twofold: First, the Pearson linear correlation of the page/posts overlap can be perceived as a falsification method for the NSD analysis conducted in the previous case study. Although it uses a different dataset (the posts that active audiences of individual pages liked on different, third-party pages), we can see that the correlation matrix creates clusters of pages with significant overlaps that significantly correspond to the clusters resulting from the NSD analysis. Like the first part of the study, we can identify two adversarial clusters and two smaller, unrelated clusters.
Second, the ČT24 example provides empirical evidence on how the active audiences of individual sites elevate particular news stories on Czech news media sites through likes. The findings indicate that specific content, particularly material related to the migration crisis, gains significant prominence on social networks through the actions of relatively small, yet coherent and active, audiences for anti-Islamic and anti-immigration movement pages on Facebook.
This elevation then influences the way news media and politicians prepare and promote their content on social network sites; shaping public debate on the crisis. Facebook closely monitors what content is trending for each page and automatically offers page editors the option to “boost” already-successful posts via paid display (promoted content). At the same time, Facebook encourages page editors to learn what kind of content their audience cares about most and repeat that style or use similar content when preparing future posts (Facebook, 2013).
Through concrete examples, we can identify which posts that alert readers to the negative consequences of immigration regularly gain significant, above average, numbers of likes on Facebook. In most cases, more than a third of these likes come from a relatively small audience: the active audience of the radical Stop Islam in the Czech Republic movement. By the same token, news aimed to bring soberer analytical information to readers is disproportionately liked by the active audience of the anti-discrimination Hate Free Culture campaign.
Given that each additional like further spreads a post to the Facebook profiles of all the “friends” of the user who liked that post, what emerges is a further solidification of “small worlds,” where similar media content circulates and similar world-views permeate.
As Bouchard and Levey (2015:2) note, the internet “may act as a facilitator and conduit for radical views online, but rarely as an all-encompassing creator of radical offline behaviour.” So far, very little is known about how individuals experience and react to the consumption of radical materials found online or about what influence it has on them (Ducol, 2015:87). Although the internet is often singled out as the key means through which individuals are radicalized, “research thus far has fallen short of unearthing the actual mechanisms through which this radicalization takes place” (Edwards and Gribbon, 2013:40). In the words of Ducol (2015:97), the internet “represents only one piece of the radicalization puzzle. Future research should pay closer attention to diachronic dynamics that may exist between online environments and ‘real world’ social settings”.
Importantly, social network sites involve real people who cannot be considered outside the socializing settings that constrain their beliefs and inform their guiding rules and daily actions in the real world (Ducol, 2015:90). Primary empirical research appears to be essential to gaining a more detailed picture of how social network sites might influence the processes of media consumption and knowledge production.
In this chapter we have presented an exploratory study on the social network sites of Czech anti-immigration and anti-Islam movements. We analysed audiences of these movements’ sites on Facebook and explored their similarities, differences, and affinities through social distance based on their fans’ likes. We used the new, formally defined, quantitative method of NSD that calculates distances between various social groups based on the intentional stances expressed by these groups’ members’ activities on Facebook. The results of NSD can be visualized in graphs or dendrograms and methods of network analysis can be applied to them. As such, NSD provides an opportunity for the distant reading of social network sites, enabling us to formally represent and analyse the structural aspects of big social data.
The methods proposed in this chapter constitute a coherent set of tools and interpretive approaches, which enable the formal representation, replication, and validation of the structural analysis of big social data and could be relatively easily adopted by other researchers in different contexts. The case study presented in this chapter could serve as an illustrative example, clarifying further possible interpretative approaches.
The main findings of this study show that, although public debate on the immigration crisis on Czech Facebook is partially highly polarized into two adversarial clusters, it is more significantly fragmented into at least four different clusters, whose audiences rarely share the same content and whose intentional stances, as manifested by Facebook likes, rarely overlap. The main findings tend to support the argument that social network sites could indeed create echo chambers and filter bubbles, thus strengthening confirmation bias (Stroud, 2010; Pariser, 2011; Flaxman et al., 2016). Nevertheless, the results are highly dependent on a specific context, i.e. the Czech migration crisis debate, and can by no means be generalized to all political communication on Facebook. Further research is needed to pinpoint specific conditions under which similar – or different – clustering occurs.
The secondary findings of this study reveal how the active audiences of individual sites elevate particular stories on Czech news media sites through likes. Specific content related to the migration crisis gains significant prominence on Czech Facebook through the actions of relatively small, yet coherent and active, audiences for anti-Islamic and anti-immigration movements. The structural aspects of the interplays between social network sites’ audiences and news media are largely neglected by both academia and policy-makers, despite their possibly significant influence on public attitudes. Beyond a theoretical framework, this chapter offers concrete methods and tools for enabling a complex structural analysis of social media sites’ audiences. Because the datasets our methods work with are publicly available, the methods and tools we propose could be used by a variety of actors (researchers, media analysts, media outlets, think tanks, governmental agencies, etc.) to support their research and decision-making processes with empirical evidence.
This chapter was partially supported by the Charles University grant project Progres 5: Životní dráhy, životní styly a kvalita života z pohledu individuální adaptace a vztahu aktérů a institucí.
Agarwal, S. (2015) “Applying Social Media Intelligence for Predicting and Identifying On-line Radicalization and Civil Unrest Oriented Threats”, Arxiv.org, Available online: (accessed 4 May 2016).
Bakshy, E., Messing, S. and Adamic, L. A. (2015) “Exposure to ideologically diverse news and opinion on Facebook”, Science 348 (6239): 1130-1132.
Basalingappa, A., Subhas, M. S. and Tapariya, R. (2015) “Understanding Likes on Facebook: An Exploratory Study”, in IV. International Conference on Communication, Media, Technology and Design Proceedings, Famagusta, North Cyprus: Eastern Mediterranean University Press.
Bode, L. (2012) Political Information 2.0: A study in political learning via social media. Unpublished dissertation, University of Wisconsin.
Bode, L. (2016) Political News in the News Feed: Learning Politics from Social Media. Mass Communication and Society, 19 (1): 24-48.
Bouchard, M. and Levey, P. (2015) “Radical and connected: An introduction” in M. Bouchard (ed.) Social Networks, Terrorism and Counter-terrorism, New York: Routledge.
Bouchard, M. and Nash, R. (2015) “Researching terrorism and counter-terrorism through a network lens” in M. Bouchard (ed.) Social Networks, Terrorism and Counter-terrorism, New York: Routledge.
boyd, d. m. and Crawford, K. (2012) “Critical Questions for Big Data”, Information, Communication & Society, 15 (5): 662-679.
boyd, d. m. and Ellison, N. B. (2007) “Social Network Sites: Definition, History, and Scholarship”, Journal of Computer-Mediated Communication, 13 (1): 210–230.
Brundidge, J. (2010) “Encountering ‘Difference’ in the Contemporary Public Sphere: The Contribution of the Internet to the Heterogeneity of Political Discussion Networks”, Journal of Communication, 60 (4): 680-700.
Cilibrasi, R. L. and Vitányi, P. M. B. (2010) “Normalized Web Distance and Word Similarity”, in N. Indurkhya and F. J. Damerau (eds.) Handbook of Natural Language Processing, Boca Raton, FL: CRC Press.
Colleoni, E., Rozza, A. and Arvidsson, A. (2014) “Echo Chamber or Public Sphere? Predicting Political Orientation and Measuring Political Homophily in Twitter Using Big Data”, Journal of Communication, 64 (2): 317-332.
Conway, M. (2012) From al-Zarqawi to al-Awlaki: The emergence and development of an online radical milieu. Counter Terrorism Exchange, 2 (4): 12–22.
Dalgaard-Nielsen, A. (2010) “Violent Radicalization in Europe: What We Know and What We Do Not Know”, Studies in Conflict & Terrorism, 33 (9): 797-814.
Dennett, D. C. (1996) The Intentional Stance, Cambridge, MA: The MIT Press.
Ducol, B. (2015) “A radical sociability: In defense of an online/offline multidimensional approach to radicalization” in M. Bouchard (ed.) Social Networks, Terrorism and Counter-terrorism, New York: Routledge.
Edwards, C. and Gribbon, L. (2013) Pathways to violent extremism in the digital era. The RUSI Journal, 158 (5): 40–47.
Ellison, G. and Fudenberg, D. (1995) Word-of-mouth communication and social learning. The Quarterly Journal of Economics, 110: 93–125.
Facebook (2013) “12 Best Practices for Media Companies Using Facebook Pages”, Facebook.com, Available online: (accessed 23 September 2016).
Facebook Audience Insights (2016) Available online: (accessed 4 May 2016).
Flaxman, S., Goel, S. and Rao, J. M. (2016) “Filter Bubbles, Echo Chambers, and Online News Consumption”, Public Opinion Quarterly, 80 (S1): 298-320.
Goffman, E. (1959) The Presentation of Self in Everyday Life, New York: Anchor Books.
Habermas, J. (1989) The Structural Transformation of the Public Sphere: An Inquiry into a Category of Bourgeois Society. Cambridge, MA: MIT Press.
Halavais, A. (2015) “Bigger Sociological Imaginations: Framing Big Social Data Theory and Methods”, Information, Communication & Society, 18 (5): 583-594.
Hegghammer, T. (2006) Terrorist recruitment and radicalization in Saudi Arabia. Middle East Policy, 13 (4): 39–60.
Huckfeldt, R., Beck, P. A., Dalton, R. J. and Levine, J. (1995) Political environments, cohesive social groups, and the communication of public opinion. American Journal of Political Science, 39: 1025–1054.
Ingram, M. (2015) “Facebook has taken over from Google as a traffic source for news”, Fortune.com, Available online: (accessed 23 September 2016).
Kosinski, M., Stillwell, D. and Graepel, T. (2013) “Private Traits and Attributes are Predictable from Digital Records of Human Behavior”, PNAS, 110 (15): 5802–5805.
Lin, D. (1998) “An Information-Theoretic Definition of Similarity”, in ICML “98 Proceedings of the Fifteenth International Conference on Machine Learning, San Francisco, CA: Morgan Kaufmann Publishers.
Manovich, L. (2011) “The Promises and the Challenges of Big Social Data”, Software Studies Initiative. Available online: (accessed 4 May 2016).
McPherson, M., Smith-Lovin, L. and Cook, J. M. (2001) “Birds of a Feather: Homophily in Social Networks”, Annual Review of Sociology, 27: 415-444.
Neumann, P. (2013) “The Trouble with Radicalization”, International Affairs, 89 (4): 873-893.
Obar, J. A. and Wildman, S. (2015) “Social Media Definition and the Governance Challenge: An Introduction to the Special Issue”, Telecommunications Policy, 39 (9): 745–750.
Pariser, E. (2011) The Filter Bubble: What the Internet Is Hiding from You, London: Penguin Press.
Parse.ly (2015) “Authority Report: The State of Tags in Digital Media”, Parsely.com, Available online: (accessed 23 September 2016).
Pelletier, M. and Horky, A. (2013) “The Anatomy of a Facebook Like: An Exploratory Study of Antecedents and Outcomes”, Annals of the Society for Marketing Advances, 25: 207-208.
Rainie, L. and Wellman, B. (2012) Networked: The New Social Operating System. Cambridge, MA: MIT Press.
Sageman, M. (ed.) (2004) Understanding Terror Networks. Philadelphia: University of Pennsylvania Press.
Sageman, M. (ed.) (2008) Leaderless Jihad: Terror Networks in the Twenty-First Century. Philadelphia: University of Pennsylvania Press.
Schmid, A. (2013) Radicalisation, De-Radicalisation, Counter-Radicalisation: A Conceptual Discussion and Literature Review. ICCT Research Paper, Hague: ICCT. Available online: (accessed 9 January 2017).
Steiner, K. and Önnerfors, A. (2017) “Introduction” in A. Önnerfors and K. Steiner (eds.) Expressions of Radicalization: Global Politics, Processes and Practices, New York: Palgrave Macmillan.
Stroud, N. J. (2010) “Polarization and Partisan Selective Exposure”, Journal of Communication, 60 (3): 556-576.
Šlerka, J. (2013) “Jak se fanoušci politických stran liší - politické strany na Facebooku (2.)”, Data Boutique, Available online: (accessed 4 May 2016).
Šlerka, J. (2016) “Polarizovaná společnost? Nikoli, je to složitější”, ReporterMagazin.cz, Available online: (accessed 23 September 2016).
Šlerka, J. and Šisler, V. (2017) “Normalized Social Distance: Quantitative Analysis of Religion-centered Gaming Pages on Social Networks” in V. Šisler, K. Radde-Antweiler and X. Zeiler (eds.) Methods for Studying Video Games and Religion, New York: Routledge (in print).
Wallace, E., Buil, I. and de Chernatony, L. (2012) “Facebook ‘Friendship’ and Brand Advocacy”, Journal of Brand Management, 20: 128–146.
Wallace, E., Buil, I., de Chernatony, L., and Hogan, M. (2014) “Who ‘Likes’ You... and Why? A Typology of Facebook Fans from ‘Fan’–atics and Self Expressives to Utilitarians and Authentics”, Journal of Advertising Research, 54 (1): 92-109.
Weimann, G. (2006) Terror on the Internet: The New Arena, The New Challenges. Washington DC: United States Institute of Peace.
Weimann, G. (2010) Terror on Facebook, Twitter and YouTube. Brown Journal of World Affairs, 16 (2): 45–54.
Appendix: Complete dataset
Daniel Herman (Minister of Culture), Martin Stropnický (Minister of Defence), Jan Veleba (p), Svobodné fórum (n), Alexandra Udženija (p), Andrej Babiš (Minister of Finance), ANO (m), Pavel Bělobrádek (Minister of Science), Blesk (n), Block Against Islam (m), Pirate Party, Milan Chovanec (Minister of the Interior), Social Democratic Party, ČT24 (n), Echo24 (n), Jiří Dienstbier (Minister for Human rights), Referendum (n), European Commission CR, European Values, Generation Identity, Hate Free Culture, Freedom and Direct Democracy, iDNES (n), Hospodářské noviny (n), Miroslav Lidinský (p), Stop Islam in Czech Republic (m), Jana Černochová (p), Marian Jurečka (p), Miroslav Kalousek (p), Christian and Democratic Union (m), Martin Konvička (p), Communist Party, Lidové noviny (n), Lubomír Zaorálek (Minister of Foreign Affairs), Michaela Marksová-Tominová (Minister of Social Affairs), NO to Brussels - National Democracy (m), Svatopluk Němeček (Minister of Health), Novinky (n), Civic Democratic Party, Parlamentní Listy (n), Petr Fiala (p), Pravý břeh (n), Miloš Zeman (President), Czech Radio - Radiožurnál (n), Reflex (n), Karla Šlechtová (Minister of Regional Development), Bohuslav Sobotka (Prime Minister), Green Party, Pavel Svoboda (p), Free Citizens' Party, Tomio Okamura (p), Tomáš Zdechovský (p), TOP 09 (m), TV Noe (n), Respekt (n), Kateřina Valachová (Minister of Education)
Note: For clarity, we have translated the Facebook pages’ names into English wherever possible (e.g. Green Party) or labelled these pages according to the following key: (n) = news media, (m) = political movement or party, (p) = individual politician. In specific cases, we provide a full description of the page in parentheses.
