1 Introduction

Internet and online social networks are diverse and almost immediate sources of information. However, platforms originally created to democratize information are currently a target of various problems that threaten democracy, such as hate speech (Castaño-Pulgarín et al. 2021) or fake news (Golob et al. 2021). In the political sphere, these practices emerge on social networks explosively and fleetingly, through the coordination of bot and particular purpose accounts, especially in times of social crisis and electoral campaigns. When a hate and disinformation campaign emerges in a coordinated and not spontaneous way, it is usually associated with the phenomenon of astroturfing. The objective is to install ideas that seem spontaneous, as if they were born from a strong relationship with the social environment. Astroturf has spread in traditional media and online social networks, especially those focused on the rapid viralization of news or ideas, such as the microblogging network Twitter (Keller et al. 2019). These communication strategies not only seek to support specific ideas but also to the detriment of others. For this, they usually resort to simplification, disinformation, and polarization of the discourse (Tucker et al. 2018; Keller et al. 2019).

Concerning strategies, we saw an update in discourse production because of new technologies that promote the new key actors, digital opinion leaders, who can or cannot, goes beyond traditional politics and dispute public agenda in another way, quicker, more emotive, and straightforward (de Roca et al. 2020; de Zúñiga et al. 2020). Precisely, it emerges as an exciting mediation possibility, part of the far-right repertory.

Although Twitter (as well as other online social networks) invests millions of dollars in the problem of detecting hate speech (Zhang and Luo 2019), this is still an open problem (Al-Hassan and Al-Dossari 2019) that affects local politics in most countries in the world (Jaki and Smedt 2019; Çöltekin 2020). Moreover, populist movements and political extremes manifested outside Twitter are replicated and reinforced in this social network (Torregrosa et al. 2020). Thus, today, governments worldwide use social networks such as Twitter to install and promote populism (Manfredi-Sánchez et al. 2021). According to Waisbord and Amado (2017), “Twitter does not present profound changes in populist presidential communication. Rather, it represents the continuation of the top-down approach of populism in public communication.”

In this article, we propose to analyze the installation of the political agenda of the Chilean far-right through Twitter as an actor from the “resistance” and the defense of the State and order. As a case study, we consider the massive protests in Chile that began in October 2019. These massive demonstrations, known as “social outbreak” or “social revolt,” gave rise to an unprecedented constitutional process, namely the formation of a Constituent Convention, the first organism democratically elected by the citizens, with representation from indigenous peoples and gender parity, to develop a proposal for a new Chilean Constitution. For the above, we analyzed five accounts of Twitter opinion leaders who reject this constituent process and the general behavior of their hundreds of thousands of followers and friends on the network. The data analysis focuses on characterizing user profiles, activity analysis, and content analysis to study the installation of speeches. We chose these three analysis approaches (user profiles, activity, and content) as far-right strategies in online social networks often rely on leaders amplifying their content across multiple satellite accounts for particular purposes, generating an echo chamber effect (Boulianne et al. 2020). Regarding the activity analysis, we distinguish between the different possible interactions on the Twitter network (Riquelme and González-Cantergiani 2016), which include posting messages (tweets), replying to a tweet (reply), sharing a tweet (retweet), mentioning another user (mention), or following a user (following). The actions of indicating “likes” to a tweet are not considered in this study, as is usual, since they are not metadata accessible from the Twitter API.

The rest of this article continues as follows. Section 2 presents recent works related to the topic of this work. Section 3 contextualizes the political background on which the analysis is focused. Here we describe the politics of the end, the political Constitution of 1980, and the constituent process during 2019–2020. Section 4 describes the analysis methodology and data collection for this work. Section 5 details data analysis and results in terms of user profiles, user activity, and user content. Section 6 discusses the findings obtained in the previous section, and Sect. 7 ends with the main conclusions and future work.

2 Related work

There are several studies about the communication strategies of the far-right on Twitter. In Crosset et al. (2018), a set of specific far-right accounts, mainly Canadian, are studied. The authors point out that these extreme ideologies manifest themselves on social media in four ways: (1) through their user profiles, (2) the content of their tweets, (3) their social interactions on the network, and (4) the use of images and memes. In our case, we start our analysis from specific accounts, but we also consider the many other accounts interacting with them, thus reinforcing the third of the four previous points. In general, we will focus on the first three manifestation ways since image analysis requires techniques still under development and beyond this article’s focus.

Through mixed methodologies, Pérez-Curiel (2020) focuses on the identification of populist discourses from three European far-right Twitter accounts: Marine Le Pen (France), Matteo Salvini (Italy), and Santiago Abascal (Spain). The authors point out that social networking services such as Twitter serve as “a stage for dissemination, interaction, mobilisation and building the personality of influencer/leaders,” also highlighting their use as strategic tools for political propaganda and installation of speeches and fake news.

In Calderón et al. (2020), the authors seek to characterize the anti-immigrant hate speech of the Spanish far-right, not from the accounts of specific politicians, but through the presence on Twitter of the political party Vox. Through content analysis and unsupervised topic modeling for topic detection, they conclude that “the hate speech against immigrants produced around Vox, and not necessarily by Vox.” The latter reinforces the idea that it is not enough to focus on particular opinion leaders but on those whose speeches impact them.

Through discourse analysis and social media analysis techniques, Åkerlund (2020) found that for the 2018 Swedish general election, far-right influencers were more active in original tweets, likes, and following relationships but less so in retweets and replies. We must relativize this result to the definition of influencers they use since it is known that, in social networks, influence can be measured in many ways (Riquelme and González-Cantergiani 2016). In Sect. 5.2, we will see how the influential users’ activity can differ depending on their interactions on the network. Åkerlund (2020) also identified a communication strategy in which a few far-right influencers are the ones who install opinions and retweet each other, being the non-influential accounts, the amplifiers or resonators that massively viralize the content. The author also identified hate speech as more negative tweets (or not necessarily negative, but loaded with sarcasm and irony) related to immigration and crime. Interestingly, the influencers were more neutral than their amplifiers, contradicting previous results indicating that the least neutral tweets tend to be the most shared. The author indicates that the latter could be a communication strategy to expand the target audience and avoid being banned from the platform.

Regarding Latin American countries, Díez-Gutiérrez et al. (2022) studied far-right political campaigns on Twitter based on hate speech from Ecuador, Argentina, Peru, and Spain. The authors employed a mixed methodology and digital ethnography, using network analysis tools, successive questionnaires, and structured discussions with ten expert academics. They verified that these campaigns did not arise spontaneously but with a destabilizing political intention, structured from organizations with marked diffusion patterns and powerful financial support. A massive presence of false accounts was evidenced, many created in the moments before the campaigns (which we will also verify in our study; see Sect. 5.1.2). These accounts acted as bots, finding identical typos in some viralized tweets. Many tweets constitute hate speech, with constantly aggressive and derogatory content. In this way, the far-right extremes ideological polarization and builds climates of tension. Nowadays, they say, minority but powerful groups, such as the Atlas Network, Hazte Oír, among others, closely related to each other and with large financial support, have specialized in this type of extremist strategies. The authors conclude that beyond a punitive approach, it is necessary to implement educational proposals to stop these strategies that threaten democratic processes.

As we can see, there is a tacit agreement in some elements that characterize the far-right movements identified from its discourses and practices. A reiterative discoursive production oriented to moralize politics by identifying common enemies from a cultural agenda, highlighting ethnic and racial minorities, feminist organizations, dissidence, sexual diversity, and left-wing militants (with communist preference). A space to hover modes of make the political question and displace management of present ways (Eribon 2018), as appropriation(s) to resignify the common in individual management of social problems. In that case, subcultures were rising, and new forms of deployment could not associate with the far-right’s emergence, which had been found in new technologies, self-mediation strategies (Cammaerts 2015), and spreading space of its ideas (Blee and Creasap 2010). This phenomenon encompass not only identity elements but also implies historiographic revisions. For instance, in South Korea (Yang 2021), the politicization through memes in Sweden (Askanius 2021), or the cover in traditional public agenda after tragedies such as Charlottesville (Atkinson 2018) to change their broadcast spaces to the Internet.

This process had been the accompaniment of algorithmic visibility of hate speech in digital platforms (Daniels 2018), which constitutes recognition pathologies elements when it hampers the far-right reason (Herzog 2021), also considered a populism paradise and an unavoidable part of the far-right action political repertories (Jacobs and Spierings 2018). It is interesting because we see continuities in strategies. No systematic study explains how platforms such as Twitter, for example, facilitate transnational far-right discourse building (Froio and Ganesh 2018). However, we can highlight how it is a co-op part of the public discussion in social networks and install discourses (Guerrero-Solé and Mas-Manchón 2022). Though the literature does not emphasize connection points, it is noticeable that inner and foreign enemy construction, migration, anticommunism, and economic discussion bring continuity, at least apparently, which deserves a more extensive study.

Indeed, several works identified Islamic phobia (Froio and Ganesh 2018) and white supremacy (Graham 2015) as enhancer enunciation spaces in social networks. Precisely, generators of politicization processes, producers of meanings and effects, which Foucault called discursive materiality.

This article focuses on the political actions of the Chilean far-right on Twitter during the beginning of the Chilean constitutional process. The most related work we have found is that of Santander et al. (2022), which analyzed the entry plebiscite campaign between November 2019 and October 2020, in which people voted for or against the drafting of a new Constitution (see Sect. 3.2). They identified a higher volume of tweets and interactions linked to the Rejection of a new Constitution campaign. However, despite this high media impact, the Approve option won by a large majority. From the analyzed data, the authors indicate that this difference is because the conservative Rejection accounts generated a lot of noise in the network but built highly homophilic subnetworks that were relatively isolated from the rest of the network. The authors associate this phenomenon with the illusion of a majority of Lerman et al. (2016). Despite the above, it is interesting to anticipate that the Rejection campaign continued unchanged, and even accentuated, during the rest of the constituent process, in such a way that in the exit plebiscite, the “Reject” option finally won with a large majority of the votes, turning the previous result. The reasons for this change are multifactorial (e.g., the drafting process of the new constitutional proposal, the change of government, the shift from optional voting to mandatory voting, etc.) and are not the focus of this work.

3 Political background

3.1 Towards a politics of the end

The politics of the end is an interpretation of Franco Berardi’s concept of phenomenology of the end, applied to the context of the social sciences (Berardi 2014, 2017). Berardi seeks to integrate communication studies, cultural studies, political philosophy, political sciences, and political psychology into a common frame, or, precisely, in an interdisciplinary approach. Phenomenology allows him to refer to the observable world and those elements that provide subjectivity. In addition, he deepens this reflection by reviewing the effects on the processes of subjectivation. In his analysis, society has a structural modification in its inner process because of a new phase of capitalism (Berardi 2017). A semiocapitalist stage that allows and promotes a new form of accumulation, in which capital is not just “money”, since we have to add symbolic capitals to the analysis (Berardi 2014). Existing is a way of accumulating because we are producing data (Sadin 2018), transferring emotions, images, and symbols that could be characterized and accumulated (Fumagalli 2015).

Society’s modification process has advanced further due to technological advances. But, the discontinuity is our experience of time, attention, sensibility, and our ability to imagine our future. Jameson (1994) said that “it is easier to imagine the end of the world than to imagine the end of capitalism”, remembering us that imagination is part of political disputes. In that sense, we assume the radicality of this approach to establish a relation of ways to appropriation.

Some similar thoughts had Jameson and Žižek (1998), mentioning that culture is a nimbus because groups have contact and observe each other as otherness. Notably, it lets us understand how danger and purity have been perceived; hate and self-affirmation as producers of sense. Based upon the preceding, no group has culture because it emerges as a nimbus as they represent, understands, and acts.

Contrary to expectations, we start with the hypothesis that far-right movements have used those strategies to dispute public agenda. We will understand the far-right as a social movement under the typology of Castelli-Gattinara and Pirro (2018). The above implies assuming their heterogeneity, their emphasis on the identity and ideological heritage to mobilize, and their highly mediated profile. Precisely, this type of approach allows us to address the problem in operational terms and differentiate the analysis of supremacist groups that do not manage to maintain collective action for a time. Likewise, it allows us to review how semiocapitalism and the accumulation of symbolic capital (like images, symbols, and signs) amplify the enunciation spaces of these groups. From these enunciation spaces emerge a narrative that, assuming the catastrophe of inhabiting a neoliberalism that naturalizes the individual management of social problems, tries to recover the grand and old narratives, such as the homeland or the social order, with an urgency that could be solved by supporting the traditional structures of knowledge. Specifically, we affirm that there is a rationality or, in Tilly and Wood (2015) words, a political repertoire, historically grounded, founded on keeping the main project of the Pinochet dictatorship: the Political Constitution of 1980.

3.2 Constituent process in Chile from 2019–2022

Political Constitution is a way to juridically regulate power exercise, that is, the way to determine margins and procedures that the State has. Also, it is a platform to manage political relations on the social space, especially in how to practice and guarantee fundamental rights.

At the time of writing this article, Chile is governed by the 1980 Political Constitution, created in September 1980 during the dictatorship of Augusto Pinochet, and ratified by a controversial national plebiscite in September of the same year. This Constitution configures an individual and commodified political convenience. Its exercise is conditioned by the neoliberal ideology, which determines citizen material conditions. In this sense, the identification of social limits that emerges from those constitutional definitions could be read from Toni Negri’s perspective of the monster (Negri 2008). It means those who are socially excluded can recognize their condition and organize their possibilities as a collective monster.

Although this Constitution has been reformed several times, different systemic and structural factors (e.g., a strong presence of the constitutional court, very high voting quotas for constitutional reform projects, the old binomial system, the interests of post-dictatorship governments, among others) had not opened the possibility of its replacement until now. However, discomfort accumulation is directly proportional to the social order promoted by the Political Constitution; therefore, the social movement (developed over several years and which exploded in October 2019) politized and visualize through political action the precariousness lives imposed by neoliberal order (Rivera et al. 2021; Bassa et al. 2020; Lorey 2018).

This study focuses on the constituent process of Chile, which started from a parliamentary agreement signed on November 15, 2019, as a result of the massive social protests that started on October 18, 2019. This agreement, signed by the country’s main political parties, considered a constitutional reform to carry out a single national plebiscite, consisting of two questions with two alternatives each:

  1. (1)

    Do you want a new Constitution?

    • Approve (“Apruebo”, in Spanish)

    • Reject (“Rechazo”, in Spanish)

  2. (2)

    What type of body should the new Constitution write?

    • Mixed Convention

    • Constitutional Convention

The Mixed Convention implied 50% of representatives chosen exclusively for the process and 50% of representatives of Congress. The Constitutional Convention implied 100% of citizens’ representatives chosen exclusively for the process. This plebiscite was held on October 25, 2020. It had the highest participation since the voluntary vote was established (50.9% of the electoral universe), and it meant a categorical victory for the Approve (78%) and Constitutional Convention (79%). This plebiscite was already in itself a historical milestone in Chile since it opened the possibility of having the first Chilean Constitution drafted by the citizens.

This constituent process generated a broad discussion of citizens at all levels, through open councils in neighborhoods, universities, and different types of organizations (Ureta et al. 2021). Mass media and online social networks have been used extensively to generate discussion around the topic. The latter is relevant, given that Chile is one of the American countries that most consumes and uses Internet services and social networks (Cardenas et al. 2018; Navarro 2020). The role of Twitter in previous Chilean social protests has already been studied. For example, (Scherman et al. 2014) show how young people in Chile used Twitter and Facebook in 2011 to coordinate mass protests within the framework of student mobilization and the “Patagonia sin represas” (Patagonia without dams) movement, the latter against the HidroAysén hydroelectric project. But the use of online social networks for socio-political purposes in Chile is transversal. In an opposite line to the previous work, Silva and Pizarro (2020) study the production and organization of post-truths by Chilean far-right movements through Facebook.

Since March 2020, due to the COVID-19 pandemic, the use of the Internet worldwide increased, further promoting its use in the constituent process in Chile. In this context, the Constitutional Convention assumed its functions on July 4, 2021, and ended its work on July 4, 2022, delivering a new constitutional text proposal. This proposal was voted on in an exit plebiscite on September 4, 2022, this time with a mandatory vote. The results were reversed, with 38.11% Approved, 61.89% Rejected, and 85.86% participation. The reasons for this drastic change are not part of the aim of this work.

4 Methodology

In this article, we will study the behavior on social networks (specifically on Twitter) for the installation of ideological discourses concerning the first years of the constituent process. In particular, we are interested in studying influential personalities, bot accounts, and particular purpose accounts related to the far-right wing. Our purpose is to show how a “politics of the end” discourse, a concept originally used by the left-wing, is configured.

The data analysis methodology used is illustrated in Fig. 1. For data collection, the Twitter API is used to collect all the tweets from the base accounts within a specified period of time (see Sect. 4.1). The collected data and metadata are stored in a non-relational database (MongoDB), due to the fact that there may be incompleteness in some attributes. Then, on these data, various quantitative analysis techniques (data aggregation or selection operations, centrality measures, statistical analysis) and qualitative analysis (text analysis) are applied to obtain and visualize results.

Fig. 1
figure 1

High-level scheme of analysis methodology used

Full size image

The quantitative analysis was carried out with the Python programming language, and the qualitative analysis with NVivo. A non-normalized General Activity was used as a centrality measure (Riquelme and González-Cantergiani 2016). Due to the amount and diversity of data obtained, the data analysis considers the following aspects, inspired on Crosset et al. (2018):

  • User profiles: Characterization of the followers and friends of the base accounts, in terms of their level of privacy (verified and protected accounts), creation dates (regarding the key dates of the constitutional process), and profile descriptions (to identify political bias and extremist ideologies). This characterization also allows us to shed light on possible influences of bot accounts related to the base accounts.

  • User activity: Activity analysis of the base accounts, their followers and friends within the network. Here, we focus exclusively on the volume of activity rather than the quality or impact of these actions. The latter is because although the actors’ centrality can be of popularity, activity, or influence (Riquelme and González-Cantergiani 2016), measuring influence (impact of the actors’ activity) is more subjective with a high presence of bots accounts. For the base accounts, we distinguish between their different types of actions (tweets, retweets, mentions, and general activity) to help define their different communication strategies in Twitter usage.

  • User content: Characterization of the communication strategies of the base accounts, based on the content of their activity within the network. For this content analysis, we focus on four themes, namely, social representations, politics of the end, verification criteria, and speech.

4.1 Dataset

This study uses public data collected from the Twitter microblogging network with the Twitter API.Footnote 1 As a starting point, five Chilean personalities accounts related to the far-right wing were considered: José Antonio Kast, deputy (2002-2018), founder and president of the Republican Party since June 2019, and presidential candidate in 2017; Sergio Melnick, minister of Augusto Pinochet during the military dictatorship (1987-1989), businessman, academic, and radio and television commentator; Teresa Marinovic Vial, mass media commentator, executive director of the Fundación Nueva Mente (New Mind Foundation); Jorge Errázuriz, economist and businessman financier of political campaigns in the sector; Gonzalo de la Carrera, businessman and media commentator. The selection criterion for this sample was convenience sampling, based on the high media profile of the five personalities and their formal political aspirations, which could empirically review whether or not their speeches have an institutional correlate beyond the far-right movement in terms of media. Their high media profile can be corroborated by their presence in the mass media and their large number of followers compared to other people on their political and ideological spectrum.

The personalities of these five accounts supported the campaign of rejecting the plebiscite for a new Constitution. Moreover, the campaign for rejection on social networks started on November 15, 2019, with a statement from the Republican Party shared by Kast through his Twitter account.Footnote 2 Paradoxically, Marinovic was a candidate for the Constitutional Convention in January 2021, being elected in May 2021 with the first majority in her sector. Sergio Melnick was elected in 2021 councilor for the Las Condes commune in Santiago of Chile, assuming in June 2021. Gonzalo de la Carrera (from now on, Carrera) was a candidate for mayor of the same commune, and was elected deputy to assume said position in March 2022. José Antonio Kast was again a presidential candidate for the presidential election in November 2021, obtaining the first majority in the first round but losing in the second round against the current president Gabriel Boric.

Based on these five accounts, which we will call the “base accounts”, the dataset contains the following data, along with its associated metadata:

  • Twitter accounts of the followers (i.e., accounts that follow them) and friends (i.e., accounts followed by them) of the base accounts (see Table 1).

  • Interactions carried out by base accounts between November 15, 2019 (beginning of the “rejection campaign”) and July 31, 2020 (three days after knowing the list of civil society organizations registered to make up the television campaign of the referendum). By interactions, the tweets posted are considered, as well as retweets and mentions carried out by the base accounts to tweets from other Twitter accounts (see Table 2).Footnote 3 We chose this date to close the data collection process since the television campaign could modify the speeches and advertising activities of the different accounts considered.

Table 1 Base accounts considered in the study (updated June 23, 2020)
Full size table
Table 2 Base accounts activity between November 15, 2019, and July 31, 2020
Full size table

All the data were collected during June and September 2020. Tweets and user accounts that have been reported and deleted (e.g., by incitement to hatred) before the collection period, are out of the sample. However, note that these data are surely a minority compared to the total obtained.

5 Data analysis and results

As illustrated in Fig. 1, in Sect. 5.1, we analyze the user profiles of the followers and friends of the base accounts to study their interest in the campaign, their ideology, and the presence of bots. Section 5.1.1 focuses on the accounts’ authenticity, Sect. 5.1.2 on their creation dates, and Sect. 5.1.3 on their profile descriptions. In Sect. 5.2, we analyze the activity of the base accounts (Sect. 5.2.1) and the followers and friends (Sect. 5.2.2) to characterize the network’s volume and interaction characteristics. Finally, in Sect. 5.3, we analyze the tweets’ content of the base accounts according to the social representations (Sect. 5.3.1) and within the context of speech analysis and politics of the end (Sect. 5.3.2).

This section is mostly descriptive (for further interpretation of the data, see the discussion in Sect. 6). The first two aspects are mainly quantitative, while the third one is more qualitative.

5.1 User profiles

The first we did was try to characterize the followers and friends of the base accounts. Based on observation, popular opinion, and formal studies such as that of Castillo et al. (2019), we expected that many of these accounts were newly created and single-purpose accounts, possibly bots. Furthermore, based on the ideas of Waisbord and Amado (2017), we suspected that a high percentage of these accounts would have a markedly far-rightist, anti-communist profile, with explicit rejection of the constituent process.

5.1.1 Accounts privacy

First, we seek to determine what percentage of the sample was verified accounts, that is, accounts whose user has decided to confirm their authenticity. Normally, verified accounts correspond to public figures (politicians, journalists, etc.) or organizations (media, companies, public services, among others). Additionally, we detect protected accounts, that is, those accounts that decide to leave their tweets visible only to their followers. It is clear that bots do not respond to profiles of public personalities. Furthermore, it is assumed that their objective is to viralize information through the network, instead of keeping “low profile”. Therefore, it is reasonable to think that neither verified nor protected accounts are bots.

The results obtained are summarized in Table 3. Unsurprisingly, verified accounts represent a small percentage of the total accounts collected. Furthermore, a percentage of 9 times higher is observed in friends than in followers. The above makes sense since base accounts are expected to follow various personalities, organizations and media with verified accounts. Looking at the base accounts separately, this difference is maintained for all cases, except for Carrera, who has a higher percentage of verified followers than verified friends. A special case is that of Melnick, who has a percentage of verified friends 23 times higher than his verified followers. In fact, at the time of data collection, Melnick follows 400 verified accounts, while 99.91% of his followers are not verified. Later we will see how many of those followers are coordinated accounts of single-purpose, which could be classified as possible bots.

Table 3 Verified and protected accounts that are followers or friends of the base accounts
Full size table

Regarding protected accounts, their number is much higher than expected, reaching more than 10% of their followers in the case of Carrera and Marinovic. Errázuriz and Kast have similar percentages among their protected followers and friends. However, Carrera’s protected followers are almost triple that of his protected friends, while for Marinovic and Melnick, they are almost double.

5.1.2 Accounts creation dates

Second, we consider the creation dates of all the followers and friends of the base accounts. Our initial assumption was that many accounts were massively created the days after October 18, 2019 (start of the 2019-2020 Chile protests), and after November 15, 2019 (day of the parliamentary agreement for the national plebiscite for a new Constitution, and the start of the “Rejection” campaign).

The number and percentage of accounts created since these dates, separated according to their relationship with the base accounts, are illustrated in Table 4. As we can expect from opinion leader accounts, there are far more new followers than new friends. More interestingly, note that 10.82% of base account followers (i.e., 65 332 followers) were created as of October 18. This very high percentage drops to 6.84% (i.e., 41 302 followers) if we consider the creations since November 15. Around 45% of these new accounts follow Melnick, Marinovic and Kast, and a much lower percentage to Carrera (17%) and Errázuriz (4.3%). However, considering the number of followers they had before October 18, 2019, Marinovic, Carrera, and Melnick are the ones who got more new followers during this period. This indicates an increase in their popularity within the Chilean right-wing. Regarding friends, the large increase in new accounts followed by Melnick (2978, corresponding to 15.94%) and Carrera (452, corresponding to 13.78%) is very striking. Thus, Melnick and Carrera reflect reciprocity in terms of the new accounts that follow them and that have followed them back.

Table 4 Followers and friends of the base accounts created from October 18, 2019, and November 15, 2019
Full size table

Our initial assumption is validated with the charts in Fig. 2, which show the evolution of account creations in the dataset over time. These charts consider the union of the followers and friends of the base accounts, since the graphs for the followers and friends separately followed a very similar behavior. Figure 2a shows the creation years of all accounts related to the base accounts from 2006 (Twitter launch year) to June 23, 2020. A greater increase is observed in 2010, the year of the presidential elections in which Sebastián Piñera, the first right-wing president of Chile after the return to democracy in 1990, was elected for the first time. Later, there is a notable decrease in 2013 and 2014, which marks the election and start of the second presidential period of Michelle Bachelet (Socialist Party). Finally, a new increase is perceived in 2017, corresponding to the new presidential candidacy of Sebastián Piñera, who was elected for his second period at the beginning of 2018. If we zoom-in in October 2019 (Fig. 2c, we can notice an evident increase in new accounts created on October 19, 2019, just one day after the massive protests in Chile began. Finally, a November 2019 view (Fig. 2e) shows an increase on November 12, the same day that the fourteen political parties opposed to the Piñera government issued a declaration in favor of a new constituent process to replace the 1980 Constitution. Figure 2b, d and f will be analyzed in Sect. 5.1.3.

Fig. 2
figure 2

Number of collected accounts (followers and friends of base accounts) by creation date (until June 23, 2020)

Full size image

5.1.3 Accounts profile descriptions

Third, we consider the information displayed by the followers and friends of the base accounts. In particular, we focus on two criteria:

  • 1st criterion Image of user profiles. While many bots or single-purpose accounts today have a highly crafted profile, not having a profile image is still considered a possible indication that the account is a bot (Efthimion et al. 2018), or at least is being used for suspicious purposes.

  • 2nd criterion Names and descriptions of user profiles. To determine if the name or description of an account corresponds to a far-right profile, we checked if it contained at least one of the following keywords:

    • “rechazo”: declaration of rejection of the constitutional Process;

    • “zurdo”: a derogatory way of referring to people on the political left;

    • “zurderío”: a derogatory way of referring to people on the political left;

    • “comunista”: communist, which in the context of the followers or friends of the base accounts, refers to a criticism of this political model or other left models (it is usually accompanied by a prefix like “anti-”);

    • “comunacho”: a derogatory way of referring to communist people;

    • “Pinochet”: refers to the figure of the dictator Augusto Pinochet (1973-1990);

    • “patriota”: patriot, a term widely used by the extreme nationalism.

Both criteria were considered on the entire sample, as well as on the selection of accounts created from October 18 and November 15, 2019. The results are summarized in Table 5. In general, there is an increase in new follower accounts without profile images. In the case of Errázuriz, although 8.91% of all his followers do not have profile images, when considering only the accounts created since October 18, the percentage increases to 15.26% (an increase of +71%). However, the most interesting results are in the friends. In the case of Marinovic, the number declines, while for Carrera, it is doubled from October 18 (+105%). In addition, as expected, none of these accounts without a profile image is a verified account.

Table 5 Percentage of followers and friends of the base accounts
Full size table

Even more interesting is the analysis of accounts created with a far-right profile, corresponding to the second criterion. Note that for all base accounts, their followers and friends with an extremist profile increase. As expected, no verified account has a far-right profile, according to this criterion. Regarding the followers, note that the highest historical base is from Carrera, with 11.71% (that is, 7439 of its 63 420 followers have a far-right profile). Since October 18, growth of between 55% and 625% (Errázuriz) has been observed. Kast has a 326% increase in extremist followers from November 15. Regarding friends, the increase is even higher, not doubling for Carrera or Melnick, but reaching incredible growth since November in the cases of Marinovic (+201%), Kast (+938%), and Errázuriz (+1568%). In raw numbers, the five base accounts add a total of 8464 friends with a far-right profile (1983 created since October 18 and 1516 since November 15), and 15 728 followers with the same profile (4261 created since 18 October and 3241 from November 15). This relation reveals a huge increase in extremist accounts of particular purpose, directly related to the base accounts and the “Reject” option of the constituent process.

The above can also be seen in the charts of Fig. 2. Figure 2b shows that the first far-right profile accounts were created in 2007. As in Fig. 2a, a strong increase is observed in 2010, for the election of the first presidential period of Sebastián Piñera. However, the extremist profile accounts do not increase massively for the start of his second term, but rather in 2019 and 2020, being in 2019 concentrated in October, November, and December. If we look at October and November 2019 (Figs. 2d and f, respectively), we notice a behavior very similar to that of Figs. 2c and e, with significant increases on October 19 (day after the start of the protests) and November 12 (statement of the opposition in favor of a constituent process).

5.2 User activity

Next, we distinguish between the activity of the five base accounts and the followers and friends of these accounts.

5.2.1 Base account activity

Table 2 shows the activity of the five base accounts for 259 days, between November 15, 2019, and July 31, 2020. By “tweets” we mean original messages that do not correspond to retweets nor mentions. The general activity is the sum of all previous actions, and it corresponds to a well-known activity centrality measure (Riquelme and González-Cantergiani 2016).

Note that the account of presidential candidate José Antonio Kast is the least active, contributing only 4.9% of the general activity of the five accounts. Despite the above, he still makes more than 9 average daily interventions on Twitter. By far, with 68% original tweets, he is the one that proportionally writes the most, without mentioning other users, even doubling the one who follows him, Teresa Marinovic (32%). Kast is also by far the one with the fewest retweets (9.2%), well below Jorge Errázuriz (34.4%). Besides, he is the only one who makes more original tweets than mentions, and more mentions than retweets. The above represents an activity behavior much more focused on speech than debate.

Sergio Melnick, on the contrary, follows a totally different behavior. He is by far the one with the highest general activity, covering 69% of all the information provided by the five base accounts, which means more than 131 average daily interventions on Twitter. His dominant activity (as with Marinovic and Carrera) is retweeting, covering almost 80% of his own interventions and 81% of retweets from the five accounts. Like Marinovic and Carrera, his activity pattern consists of making more retweets than original tweets, and more original tweets than mentions. Thus, Melnick’s behavior on the network is much more focused on the content or opinion viralization than on speech, paying much less attention to the possibility of debate.

Teresa Marinovic follows the same pattern as Melnick in her Twitter activity. However, she is the second least active of the five (7.4% of the total general activity, with 14 average daily interventions). Unlike Melnick and Kast, her behavior is a bit more heterogeneous, being the second with the highest percentage of original tweets (32.1%), after Kast, and the second with the highest number of mentions (23.4%), after Errázuriz. Therefore, although content viralization predominates in her communication strategy, she also focuses on speech, without neglecting the possibility of debate.

The case of Jorge Errázuriz is also somewhat different from the previous ones. Covering 10% of the general activity of the five accounts, with 19 average daily interventions, he is the only one in which the debate attitude predominates (covering 47% of the mentions of all accounts), following the content viralization (although he is the second with the lowest percentage of retweets, only surpassing Kast) and finally the speech. Besides, he is the one who presents the most heterogeneous behavior.

Finally, Carrera has a general activity only slightly lower than Errázuriz (8.7% of the general activity of the five, with 17 average daily interventions), and presents a pattern of behavior similar to Marinovic and Melnick. However, his behavior is less heterogeneous, being the second one with the highest percentage of retweets (68%), the penultimate in terms of mentions (12.9%), and the last one in terms of original tweets (19.1%). This behavior is similar to Melnick, who prioritizes the content viralization from other users with a similar ideology rather than the speech or debate.

5.2.2 Followers and friends activity

We decided to analyze the activity of the followers and friends of all the base accounts jointly, since separately, they present several similarities, not only in terms of matches (around 260 thousand user accounts) but also in terms of their behavior. Our sample considers 603 879 followers, 129 853 friends, and 683 325 accounts without considering duplicates. We also distinguish between accounts with far-right profiles, according to the stated in Sect. 5.1.3.

Figure 3 shows the general activity of the followers and friends of the base accounts, ordered from highest to lowest. We already explained the general activity measure in Sect. 5.2.1, and we use it in Table 2 for the base accounts. Note that the distributions behave according to the Pareto Principle, i.e., around 80% of the total network activity comes from 20% of its users (Jinbo and Hongbo 2019). On the other hand, the most active accounts in Fig. 3a generally correspond to press media, while the most active ones in Fig. 3b are all personal accounts.

Fig. 3
figure 3

General activity of the followers and friends of the base accounts

Full size image

To complement our analysis, we decided to correlate the shares of likes with the general activity of the accounts. Since we cannot assume a normal data distribution, we use Spearman’s correlation coefficient. The results are shown in Table 6. All these results are statistically significant, with a p-value \(<0.001\).

Table 6 Correlations between likes and general activity (p-value \(<0.001\))
Full size table

According to Table 6, we observe that the correlations obtained are high, especially for the followers and the entire sample, without filtering by extremist profiles. The only moderate-high correlation is that of accumulated friends with a far-right profile. A high correlation can be interpreted as that the action of putting likes to tweets is a faithful indicator of the overall user activity. Indeed, on Twitter, users’ retweets are usually accompanied by a “like”. Interestingly, note that in this case, the correlations decrease in all cases for the far-right profiles. Moreover, the correlations are also lower for friends than for followers (remember from Table 5 that there is a higher proportion of friends than followers with a far-right profile). The above means that in the far-right profile accounts, the “likes” are less representative of their global activity: they seem to be more interested in viralizing content and opinions, through posts, replies, or retweets, rather than in expressing their approval of a certain type of content. This emphasis on content viralization over content approval is also a behavior that could be associated with particular purpose accounts.

5.3 User content

Tweets’ content illustrates four major themes expressing far-right political collective repertory. These themes are summarized in four concepts shown in Table 7. These themes include: social representations, considering how they perceived otherness; politics of the end, for understanding radicality strategies; verification criteria, if the use or not, ideological elements; and speech, highlighting its performance. Indicators of the topics are the following: homeland, international relations, power relations, and terrorism; catastrophe, urgency, management, and repression; ideology, (post)hegemony; structure, political articulation, said, and unsaid; respectively.

Figure 4a-e illustrates data by automatized word clouds visualization using NVivo software, considering all that the five base accounts say separately. Figure 4f also shows a consolidated cloud with all the tweets. The tweets considered also include the retweets and mentions made by the base accounts since both actions consider at a discursive level a seizure of the original tweet message. Then, we use the analysis matrix to problematize tweets content by major themes. This tries to integrate quantitative and qualitative strategies remembering mediations, precisely the description of the place where constraints delimit and shape social relations, materiality and cultural expressiveness (Barbero 1993).

Table 7 User contents matrix for qualitative data analysis
Full size table
Fig. 4
figure 4

Word clouds of base accounts and accumulated

Full size image

5.3.1 Social representations

According to the social representation theory from (de Rosa, 2012, Ch. 2, pp. 77–88), we can assume three dimensions of their approach: cognitive, relational, and practical. We orient these dimensions to a perspective that assumes the world as a “between” of people because we create language, we represent things, and then, we exist, as Arendt (2019) says.

Applying this to the data, we observe the Homeland indicator in all clouds. The use of the word “Chile” refers to the declaration of a project, creating the representation of the “politicians” (políticos) and “political” defenders, the only ones who can defend the “Constitution” (Constitución) and protect the country from “communist” (comunista, @pcdechile) “violence” (violencia).

Regarding International Relations, the tendency is clear. Chilean far-right identify “world” “communist” (otros, Estados, Gobiernos) referring tacitly to Maduro, Venezuela, and the United Nations, as responsible for protests and smear campaigns from human rights, respectively. Even so, when the discussion comes from Power Relations, tweets are critical of the “government” (Gobierno), which is criticized for being too flexible with the left-wing. The same occurs for Terrorism, wherein an internal lecture of the present, crimes come with “violence”.

5.3.2 Politics of the end and speech analysis

To talk about the politics of the end, we need to assume the catastrophe. Or, remembering Benjamin (2019), that history and progress always come with the force of destruction to create something new. Interestingly enough, in this case, we do not need to criticize capitalism and their possibilities because Catastrophe is a strategy to defend the current social order. In that sense, instead of talking about the causes of Chilean’s outbreak, forgetting the causes and assuming that the political process is spontaneous, they focus their analysis on “violence” (violencia) to perform political dissent are part of the discourses from tweets. Besides, the sense of Urgency is appreciated in words like “now” (ahora).

When the discussion goes beyond strategy, Management and Repression become protagonists. Mentions of “state” (estado), “government” (of) “right” (derecha) or “Piñera”’s (President of Chile in 2018-2022) lack of leadership to confront the social mobilizations are recurrent. Additionally, the grateful attitude to the police, “carabineros”, for stand by the Republic is sustained.

The discussion of what is ideological or not in politics is as old as the separation between politicians and scientists to rule societies. But, we can avoid how ideology let domination, made consensus, and perform hegemony. Gramsci (2015) proposes a path to legitimize social order without coercion. Particularly, we found “left” (izquierda), “communist” and “Vallejo” (from Camila Vallejo, a member of the Parliament), “Front” (from Frente Amplio, a left-center coalition) identified as enemies of the order. The word “politicians” (políticos) is constant and also appears associated with negative critics, even when most of the base accounts were or aspire to get a political position (see Fig. 4a for the case of Kast, the current presidential candidate from the far-right wing, or Fig. 4e for Carrera, currently a sitting deputy).

Tweets’ speech has a clear Structure. The base accounts assume that their actions are necessary to defend the “constitution” for the far-“left”, who has locked political possibilities with “violence”, and “offenders”. The Political Articulation of their speeches is maintained through the relation “violence”, “left”, and the imposition of the constitutional Process. Precisely, they say that the “social” “agreement” (acuerdo) for a new Constitution only was achieved because of the high level of conflict (crisis), and thus the process is illegitimate. Differing from the Unsaid, i.e., that they agree with the current Constitution and the silence about human rights violations since October 2019. A tweet that synthesizes the discursive essence presented here is the following:

“We are the resistance. We are many. We are more. In silence, and with humility. Keep working for Chile! #YoRechazo [#IReject]” @joseantoniokast (Jan 9, 2020)Footnote 4

6 Discussion

The purpose of this study was to examine the “politics of the end” as a repertoire of political action of the Chilean far-right in the context of coordinated activities on Twitter during the beginning of the Chilean constituent process. The study found that the five chosen opinion leaders emerge as power figures.

Next, in Sect. 6.1, we will discuss the activity and action strategies of the base accounts and their followers and friends. In Sect. 6.2, we conduct a content analysis of the installation of speeches given by the base accounts, amplified by their followers, many of whom act as echo chambers.

6.1 Discussion about activity and action strategies

It is important to highlight the far-right’s almost immediate reaction and coordination capacity in the face of eventual political crises. In the context of this study, it was faster than expected, managing to organize itself as of October 18 and November 15 (see Table 5 and Fig. 2). This massive creation of particular purpose accounts does not seem to respond to improvisation nor a spontaneous event.

The base accounts, corresponding to opinion leaders of the Chilean far-right, increased their followers substantially as of the social outbreak of October 2019, with a significant number of new accounts with marked profiles of far-right militancy (Pinochetist, anti-communist, in favor of the Reject option). Proportionally, the accounts that grew the most in followers were those of Marinovic, Carrera, and Melnick, who had a significant media presence on social networks. Many of these accounts are for particular purposes (see Sect. 5.1.2), presumably created to function as echo chambers in political campaigns for the Reject option and, more in general, to maintain the status quo through an ultra-conservative discourse (see Sect. 5.1.3). Interestingly, many of these accounts were created during the presidential campaigns of Sebastián Piñera (the first right-wing president after the return to democracy). However, the far-right profiles increased, especially in 2019. From all of the above, it is presumed that the right wing has organized various campaigns on the Twitter network, made up of multiple accounts with bot behaviors, acting as amplifiers of the speeches spread by specific political leaders.

Notwithstanding the preceding, these political actors’ (high) activity may differ in the same event or campaign, complementing each other. Thus, Sect. 5.2.1 shows that some actors are more focused on the installation of discourses (Kast, Marinovic), others on the viralization of content (Melnick, Carrera), and others on the debate (Errázuriz, Marinovic). On the other hand, the activity of the followers and friends of the base accounts (see Sect. 5.2.2) follows the Pareto principle. However, the accounts with extremist ideologies do not seem as interested in expressing a liking for other messages as in the single viralization of content. Hence, the most extreme accounts around the base accounts act more like bots and echo chambers.

6.2 Discussion about content analysis

Language appears when it has the power to be a repeat and social update (Foucault 1971). Their strength of repetition comes from legitimacy places, which means that it comes from someone’s position when the discourse is delivered. We consider five relevant subjects in the Chilean far-right who can influence the constituent process, including 603 879 followers (without duplicates) and almost 50 000 tweets. This study shows a potential field to evidence repetition, transformation, reactivation, and political effects statements. From this study we can ratify the qualitative findings of Silva and Pizarro (2020) that suggests Chilean far-right performs a political repertoire on Internet based on their resistance to the left-wing.

Promote the idea of foreign intervention and hide from the discussion of precarity, inequality, and social rights are part of the action repertoire of Chilean far-right. Assuming the catastrophe’s urgency produces a discontinuity in similar studies (Berardi 2014, 2017), focused on descriptive content analysis and forgetting the discourses’ political articulation. Political Sociology, Psychology, Philosophy, and Communication have discussed many potential paths and political articulation possibilities. However, the politics of the end highlights in to assume the catastrophe produced by the capital as a political possibility. Their radicality is to think from disaster and survival, and a politics that problematizes life concerning the current process of capital accumulation (Rivera et al. 2021).

Applying the matrix from Forscher and Kteily (2019), just in an experimental way, to far-right tweets in the Chilean case, we found some similarities. Firstly, the main concern about the future (what we called catastrophe), emphasizing economic expectations for the advance of mobilizations and a potential “Chilezuela” (considering the possibility to replicate the Chávez-Maduro process in Chile). We found a particularity in this study from external enemies because of the social composition. In Chile, the apparently depoliticized argument, hidden by a moral framework (Bouvier 2020), avoids the quid of the discussion: the political radicalism and the denial of the socio-structural elements that underlie the social protest.

The use of “others” (studied in Japan referring to East Asian people in a sociological approach to emotions (Asahina 2019)) reveals a dissociation in sociability with an ontological perspective considering what we are and what we stand for. Precisely, in our study, “others” are communists and terrorists who want to steal “the progress” from the last thirty years in Chile. We found a similarity with Erdogan-Ozturk and Isik-Guler (2020) and their study in Turkey with refugees on Twitter. In Chile, there is no talk of the presence of refugees, but the common internal enemy identified in our research is the left and its derivatives (communist, Venezuelan, Cubans and Mapuches, all of them “terrorists”), which are signed in Twitter as the reason of the political crisis.

However, the big difference, considering our data, is confidence in the institution. Chilean far-right does not have high confidence in alternative media to express ideas; they had an institutional tendency. We found many references to mainstream media like La Tercera o El Mercurio, who often showed our five opinion leaders in their tweets. The above, considering the high presence of conservative values, because of property, in Chilean mainstream media.

Furthermore, considering the structure of discourses and the unsaid, Chilean far-right has a lot to analyze. If we look for similar studies with alternative frames like Japan (Schäfer et al. 2017), we found opportunities to dispute common sense from politics discredit in digital space, considering its potential and non-specificity. In Weber (2019) words, a length of the void that lets charismatic leadership appears to dispute and constitute legitimacy. The extremist discourses identified here reterritorialize (Lazzarato 2019) symbolic values like homeland, social order, and political stability by a continuous stream of digital data.

Similarly, within the digital environment’s technological transition, we had come to a defining point of dissociation between empathy and social link (Berardi 2014, 2017). Increasing exposure to streaming and the exponential growth in data volume limits our capacity to elaborate critical analysis, saturating our sensibility and time of attention. If attention is an economy (Fumagalli 2015), the substitution of political will by a biosocial design inserts automatically and algorithmically answers in our perception, imagination, and desire. In this study, we have shown the Chilean far-right as a possibility to analyze. It has been studied in other cases like Belgium (van Erkel and Aelst 2020), considering how information (data in our case) is overload and attempted in political learning. In that sense, questions have to be in their effects.

The power of Twitter bots attempting to shape or disrupt public opinion should not be ignored (Mintal and Vancel 2019), especially considering their possibility to suppress the dissent (Neyazi 2019), polarize available agenda constituting “cyber-armies” (Wang et al. 2020), and become a concern from democracies (Rheault and Musulan 2020). In this study, although we cannot assure the presence of bots within the sample of accounts collected, there is evidence of active coordination of accounts created with extremist profiles and a particular purpose. Chilean far-right called several mobilizations and “tuitazos” (Twitter’s reunion to make a trending topic) using coordinated strategies in online social networks for the “Reject” campaign. After their clear defeat from almost 80% for the “Approve” option in the entry plebiscite, these coordinations shifted to discredit the Constitutional Convention and other political issues, such as the presidential elections. In the presidential elections, they were unsuccessful, although Kast won the first round with the first majority, being defeated in the second round by Gabriel Boric. However, in the exit plebiscite, the result was reversed, winning the “Reject” option with a large majority.

Maybe far-right, and not only Chile as we have review, are the political group that learned better what algorithmic governance is. The intermediate space and the production of (a) political rationality type (Rivera et al. 2021) that relies on the automated collection and extensive database analysis, which can model or anticipate social behavior (Rouvroy and Berns 2013). Its appropriation of these spaces, in digital space itself and for the self-mediation potential (Cammaerts 2015) is a new phenomenon for social sciences and a problem to study.

7 Conclusions and future work

This article began with the hypothesis that Chilean far-right wing movements use strategies of the politics of the end as a political repertoire to dispute the political agenda. From social network analysis and the analysis of quantitative and qualitative data applied to the content spread by 600 thousand Twitter accounts in the context of the 2019-2020 constituent process, we have been able to advance towards the verification of this hypothesis. Indeed, for the dataset considered, it is evident how supposedly left strategies dialogue for the political action of the far-right. It is interesting how extremist ideologies can generate trends in social networks, installing discourses on the public agenda, beyond their limited parliamentary representation. It also highlights the great response capacity of the far-right to coordinate and install speeches on the social networks and the different profiles that opinion leaders can assume: discursive, amplifiers, and debaters.

From a methodological and content point of view, political analysis studies in social networks are usually mostly descriptive, or they are not very explainable when they are focused on the use of machine learning techniques. Therefore, we believe that this work can shed light on new ways to approach similar studies. The focus of this work has been markedly interdisciplinary, using computational tools for social network analysis together with a discussion in a philosophical and sociological way. In this sense, it has been possible to observe that the far-right operates the difference between “them” and “us” as the articulating axis of their speeches and adoption of warlike identification strategies, with the urgency of the catastrophe.

In this work, we have considered studies from Japan (Asahina 2019; Schäfer et al. 2017; Mintal and Vancel 2019), Taiwan (Wang et al. 2020), India (Neyazi 2019), Turkey (Erdogan-Ozturk and Isik-Guler 2020; Çöltekin 2020), Germany (Jaki and Smedt 2019), United States (Forscher and Kteily 2019), Canada (Bouvier 2020; Rheault and Musulan 2020), and Latin America (Castillo et al. 2019; Waisbord and Amado 2017) and we have noticed how homeland as an abstract concept is territorialized (Lazzarato 2017) in each territory to produce identification. If globalization transforms identities because of semiocapitalism, the far-right wants to recover its concept to perform political action.

A possible line of future work would be to identify how these strategies in online social networks and their derived public discussions affect the public agenda. Additionally, from the point of view of social network analysis, it would also be interesting to know the impact of the content generated by opinion leaders of extremist ideologies within the network. For instance, the impact of the opinion leaders could be analyzed in terms of the spread of their influence within the network.

Finally, the dataset could be useful for other types of analysis, such as sentiment analysis or discourse analysis. From this work, we have verified the recycling of old accounts for new campaigns that are more ideologically extreme. As future work, it would be interesting to follow up on the dataset obtained to analyze how the echo chambers behave once the event to which they were dedicated has ended. This approach could be applied to datasets from other countries derived from previous research.