Journalism in Democracy: A Discourse Analysis of Twitter Posts on the Ferrerasgate Scandal

This research analyses the discourse on Twitter surrounding the “Ferrerasgate” scandal involving the Spanish journalist Antonio García Ferreras, director and host of the television show Al Rojo Vivo (La Sexta, Spanish TV channel). It examines the main object of criticism, the tone of the discourse, the argumentation made by users, as well as the existence of hate in their rhetoric. The tweets included in the study’s sample ( N = 2,846), posted between 5 and 15 July 2022 and extracted on 16 July 2022, were examined in two complementary phases. The first entailed a quantitative content analysis of the messages and the second analysed whether hate speech was found in the sample as a whole. The Sketch Engine tool was used to determine whether “crypto hate speech” existed in the sample as a whole, and to whom it was targeted. The results reveal that “Ferrerasgate” sparked a debate that spilt over into journalism across the board, calling into ques‐ tion the media’s role in a democracy. The most prominent arguments were the condemnation of misinformation, lack of independence, and absence of professionalism in the journalism sector. It should be noted that most of the messages were destructive in tone; hate was found in the tweets analysed, although these did not represent a high percentage in relation to the total sample.


Introduction
The debate on the status of journalism is not new, but recent events seem to have strengthened and bolstered the profession. In the digital age, where information spreads around the world at breakneck speed, the new technological context, the different consumer tastes and habits (Carlón & Scolari, 2009), the phenomenon of disinformation (Alonso, 2019), and the changes in the ethical principles that once underpinned journalism (Rodrigo-Alsina & Cerqueira, 2019), not to mention specific events such as the Covid-19 pandemic Reguero-Sanz et al., 2021) or Ukraine war (Pavlik, 2022) have placed journalism at a historical crossroads. Twitter has become an established professional tool for journalists, whether used as a source of information, a space for conversation, or a forum to observe trends of opinion (Molyneux & McGregor, 2022). However, in parallel to these new opportunities provided by Twitter, it has also intensified a number of negative phenomena that have historically affected good practice in journalism. In particular, misinformation and the causing of hoaxes to go viral. Several studies have examined the link between Twitter and the spread of fake news, for example, in the case of Donald Trump's election victory or the Brexit referendum (Cabezuelo & Manfredi, 2019). Similarly, Vila Márquez and Arce García (2019) applied big data techniques to analyse the process of dissemination of fake news on Twitter and concluded that there were organised structures for the re-dissemination of this type of fake content. Likewise, Sivasankari and Vadivu (2022) monitored how hoaxes were reproduced using graph theory. In response to the increasing amount of fake news on Twitter and the impossibility of confirming the veracity of the content, some initiatives have been launched to combat fake news through algorithmic methods such as deep learning (Sakshini & Bhavna, 2022). The defence of the truth has also been advocated as the only solution to combat journalism's lack of credibility (Sixto-García et al., 2021).
Twitter has also provided new avenues for sensationalism that have even affected how news agencies use their profiles (Brown et al., 2016). In fact, due to the specific formal peculiarities-a restricted number of characters-and its transmedia properties, Twitter has encouraged dishonourable business behaviours such as clickbait, closely linked to sensationalism (Bazaco et al., 2019). The Covid-19 pandemic was also a vector that affected the media's sensationalist use of this social network (Sánchez-Figuera et al., 2021). The spectacularisation of information, a phenomenon directly related to tabloid journalism that has been a trend in Spain since the arrival of private television (García-Avilés et al., 2022), has also been encouraged by Twitter. For example, the use of memes by users has served to further increase the already existing toxic sphere in relation to journalistic content .
The characteristics of Twitter have somewhat accentuated these major problems and created a fracture between society and the media. Not only is there a media credibility crisis (Barrios-Rubio & Gutiérrez-García, 2022), but we are seeing the emergence of personal and professional finger-pointing. Miller (2021a) noticed this trend and developed a theory related to it, which demonstrated that the harassment experienced by journalists on social media directly affects their professional performance. This has led to them avoiding interviews with certain leaders and has had an impact on their personal freedom, which has ultimately resulted in their becoming less active on social media (Miller, 2021b). The type of journalistic specialism is also an influential factor in the violent communication received by media professionals, with politics and gender being the two profiles that most often experience aggressive behaviour from social media users (Kim & Shin, 2022).
Verbal aggressiveness is directly related to the ideological polarisation of the 2010s and even to hate speech. Journalists have not been considered to be a marginalised group and therefore the target of this type of rhetoric, as hate speech was only deemed to affect vulnerable groups or those at risk of social exclusion (MacAvaney et al., 2019). However, recent research has identified a type of discourse that targets other individuals or groups (such as journalists) who are not discriminated against because of their economic status, gender, race, or religion (Shin et al., 2020). This phe-nomenon is defined as "socially unacceptable discourse" (Vehovar & Jontes, 2021) or, more precisely, as "crypto hate speech" (Herrero-Izquierdo et al., 2022). This term refers to rhetoric found in the social media environment which is semantically constructed in a similar way to traditional hate speech, although it cannot be considered as such because it is not targeted at vulnerable groups. This research underlines the interest in this type of discourse that has been perceived at the political level, where much of the violent communication on social media is found.
This theoretical framework contextualises the initial purpose of this research, that is, to examine the discourse produced by the social media audience on a specific media-political scandal. This article aims to investigate Twitter users' reactions to this event, as part of a rising wave of polarisation (Weismueller et al., 2022). The focus is the case popularly known as "Ferrerasgate," concerning the manipulation of information about Pablo Iglesias, former vice-president of the Spanish government and founder of the left-wing party Unidas Podemos. This was uncovered in July 2022 as a result of the recording of a conversation between police commissioners José Luis Olivera, José Manuel Villarejo, Atresmedia executive Mauricio Casals, multiple business owner Adrián de la Joya and journalist, director of La Sexta TV channel and host of the programme Al Rojo Vivo, Antonio García Ferreras. In May 2016, news about an alleged bank account belonging to Pablo Iglesias in the Grenadine Islands, a tax haven, and his relations with the Venezuelan government of Nicolás Maduro were broadcast in Al Rojo Vivo. The information would prove to be false after having been first published in Okdiario, the newspaper directed by Eduardo Inda. However, the controversy arose because, despite giving little credibility to the news, the director of La Sexta (Spanish private television channel) decided to broadcast it: "I told him: 'Eduardo [Inda, director of OKDiario], this is very serious. I'm going with it, but this is very delicate and too crude"' ("Ferreras, sobre los audios," 2022).
Besides other issues mentioned in the audio recording, such as political conspiracies to sink the career of Juan Carlos Monedero (co-founder of Pablo Iglesias' party), it notably sparked a heated discussion in the media industry and, above all, on social media platforms. Specifically, the tweets posted on this topic in July 2022 generated a significant flow of information and opinions on the subject, including more-or-less critical views. What first appeared to have caused yet another flurry of comments on Twitter gave rise to an extensive digital discourse that displayed different ideologies.
Once again, Twitter became the preferred arena for dialectical confrontation (Burnap & Williams, 2015), a framework that was representative of the existing political polarisation (Friedolin & Volfovsky, 2018;Urman, 2020) and a repository for a form of hostile communication that often bordered on hate speech (Gómez-García et al., 2021). "Ferrerasgate" exposed the dilemma of credibility, disinformation, diligence in the use of sources, as well as respect for the truth in journalism: "Villarejo's audio recordings of Ferreras uncover the origin of the news about Pablo Iglesias' fake bank account" (Precedo, 2022); "journalism, a profession in palliative care" (Cacho, 2022); "FerrerasGate: what to do with the 'sewer' media" (Alfonso, 2022); "Journalism is in need of a catharsis" (Barceló, 2022).

Objectives and Method
Given the relationship between Twitter discourse and journalism, the hostile environment faced by news professionals and the specific case in question, the general objective of this research is to analyse the argumentative criticism generated by Twitter users in relation to the "Ferrerasgate" scandal. This general objective can be broken down into three objectives (O): O1: To examine who the main object of criticism was in the "FerrerasGate" Twitter conversation. That is, to ascertain if the messages focused on the main journalist involved (Ferreras) were extrapolated to his colleagues in the programme, in the television channel, or even to the rest of the profession.
O2: To identify what type of argumentation predominated in the tweets. To analyse whether users focused on debates about the journalism profession or whether their criticism was solely targeted at the characteristics of this specific case.
O3: To explore whether the messages posted had a constructive or a destructive purpose. That is, to examine whether they contribute to the improvement of journalistic practice or whether they are intentionally hurtful messages. This will allow us to determine whether hate speech was perceived in the tweets analysed.
Three starting hypotheses were formulated that were consistent with the three specific objectives outlined above: H1: The discourse on Twitter was extrapolated to the entire journalism profession, despite the references necessarily made to the case study at hand. H2: The criticism in the tweets was dominated by the argument against disinformation since the scandal began with a report revealing that the director of the programme Al Rojo Vivo had authorised news stories despite his doubts about their credibility.
H3: This scandal resulted in an aggressive rhetoric against journalism in general on Twitter that was specifically aimed at the protagonist of the scandal, which manifested itself as crypto hate speech.
In order to verify or refute the proposed hypotheses, this study used quantitative content analysis, a technique used in previous research to obtain an objective and systematic understanding of the content of the messages analysed (Riffe et al., 2019). The study sample was selected from tweets containing the handle @alrojovivo posted on Twitter between 5 and 15 July 2022 (N = 2,846).
The tweets were compiled using text-mining techniques, which allow information to be extracted from big data with minimal human interference (Lin et al., 2016). RStudio free software (version 1.4.1103) and VOSON Dash library data were used for data collection. A link was established using the application programming interface for Twitter developers with an account registered for academic research. This is a data acquisition method tested in recent research (for example, see Arce & Vila, 2021).
The messages were examined in two complementary phases. In the first phase, a detailed analysis was made of each tweet in the sample. Three researchers undertook the coding work and randomly tested it using Krippendorff's Alpha coefficient to calculate reliability. A pre-test with tweets posted on different dates and by different users provided a reliability coefficient of 0.855 in the overall computation, confirming that the three coders followed the same criteria. The variables analysed were divided into two groups: form-based and content-based.
There were six form-based variables: • Followers of the tweet's author.
• Date when the tweet was posted.
• Attachments to the text message. Memes were categorised based on the ideas of Paz et al. (2021). Based on the purpose of the message, the authors classified this content into two macrocategories: positive and negative. Within the latter, they established different purposes of a visual montage: ridicule, criticism with hate, and hate speech. The memes were analysed due to the role they play in fragmenting digital audiences and creating an emotional political narrative. Macro image memes and the most popular meme were selected, whereas those that had no image, video, or GIF were excluded, following the formulation by Zenner and Geeraerts (2018).
The main aim of this research was realised by studying the content of the tweets. The three variables analysed relate to each of the specific objectives of the study. The first variable was the Target of Criticism: (a) The journalism profession-those tweets that specifically referred to journalism practice as a whole; (b) TV channel-messages that mentioned La Sexta; (c) TV show-references to the name Al Rojo Vivo; (d) Host-the key player in the scandal, Antonio García Ferreras; (e) Other-any other individual mentioned in the tweet (whether they were journalists or not). More than one option was checked if the message was targeted at several people and/or groups. Following the corresponding analysis, saturation sampling was used in order to identify the various categories that make up this variable.
The second analysed variable was the Main Topic. Seven subcategories were included to classify each of the messages according to the reasoning and the main topic to which they referred, with saturation sampling also used in this case. The seven subcategories were: (a) Lack of Independence-this section categorised any tweet related to corruption, "sewers," and the controversial relationship between politics and journalism, where the latter depends on the power elites in society; (b) Professional Protectionism-this included examples of messages accusing either La Sexta or the television programme Al Rojo Vivo of covering up "Ferrerasgate," and failing to thoroughly question the information about the scandal to avoid harming the reputation of the channel, its team, and/or its professional and commercial interests; (c) Lack of Professionalismmessages demanding the resignation of the main individual involved in the scandal, Antonio García Ferreras, and those calling for the show to be cancelled or for the television channel to be shut down, includeding users' claims concerning the lack of journalistic ethics regarding the scandal; (d) Disinformation-this category included arguments referring to the lack of credibility of the target of criticism and tweets related to the alleged manipulation engaged in and presumed disinformation spread by the media; (e) Partiality-this section included tweets in which users perceived imbalanced news coverage, or in other words, where one of the key actors involved could voice their views, while another was silenced or not allowed to retort; (f) Groundless Arguments-this included messages with no reasoning whatsoever, that is, tweets in which the user only hurled insults against the target of criticism, additionally the results obtained in this section determined the analysis phase in which hate speech in those tweets was assessed; (g) Otherslastly, this group included messages which, for one reason or another, could not be classified into one of the previous sections (unrelated to the subject of study, decontextualised, etc.).
The third variable analysed was Purpose, which was divided into two subcategories based on the study by Ksiazek (2016) in which a conversation on Twitter is divided into terms of two large blocks: civility and hostility. This idea has been adapted to the present research by classifying it as: (a) Destructive-this variable was considered when a tweet contained empty reasoning and/or insults, or if sought to harm or ridicule the target of criticism, and these posts were found to convey highly emotional messages and feelings linked to rage, anger, disillusionment, and disappointment, among others; (b) Constructive-this variable was taken into account when a tweet contained logical and coherent arguments rather than emotional ones, as such, the post might contain objective and verifiable facts reaffirming and supporting the user's view.
The second phase of this study, which determined the extent to which crypto hate speech was used in the sample, entailed classifying the tweets according to the "groundless arguments" variable which, by their nature, also form part of the "destructive" category. Subsequently, the discourse of this type of messages was analysed using the Sketch Engine software. This is capable of automatically generating lists of words which, are placed within the same semantic category according to their use (percentage of shared collocations), avoiding any type of bias on the part of the researcher. In other words, this list is produced according to the context in which the words appear in the selected text corpus. In this way, we could find out which were the most repeated terms on which the hostile discourse was focused. The "thesaurus" tool made it possible to locate words with similar meanings or which appear in the same context.

Object of Criticism
This study comprised a sample of tweets collected from various Twitter accounts with varying degrees of impact (Table 1). Nearly all of the 10 most influential accounts that tweeted about this episode post political content regularly.
All the accounts that posted political messages, such as @protestona1 and @AnitaBotwin, were prone to use Twitter for political activism, leaning towards the left. These accounts act as a mouthpiece for journalists closely identified with that side of the political spectrum, such as Antonio Maestre, Julián Macías Tovar, and Rosa María Artal. The @HPodemita handle made its opinion on the debate known by changing its header photo to one of Antonio García Ferreras superimposed with the word "manipulation" emblazoned on it (Figure 1). The rest of the Twitter accounts analysed in this study that do not regularly publish political content are focused on their users' professions, such as graphic design or photography.
These prominent Twitter accounts and the other less influential users involved other users in their messages by using the "mentions" feature. In addition to the handles of those directly involved in this newsworthy incident (Al Rojo Vivo, La Sexta, Pablo Iglesias, and Eduardo Inda), many of the channel's journalists were also mentioned, including Ana Pastor (55), Rodrigo Blázquez (80), Joaquín Castellón (18), Cristina Pardo (four), Jordi Évole (five), Iñaki López (three) and Helena Resano (one). This indicated that the news manipulation scandal affected the credibility of the entire La Sexta channel. However,  (two) and Antonio Naranjo (one). Moreover, other television or radio programmes or other media also surfaced in this social media discourse: El País (four), Público (four), El Plural (three), En Jake (three), The Guardian (three), Herrera en COPE (three), Cuatro al Día (two), El Programa de Ana Rosa (two), Espejo Público (two), La Hora de TVE (two) and El Mundo (one). This finding clearly demonstrates that the publication of the Antonio García Ferreras audio recordings sparked a meta-media dialogue involving communication industry professionals and companies (Figure 2).

Type of Argument
The tone of the tweets was highly critical and most of them questioned the work of Antonio García Ferreras (95.33%). Moreover, it affected the credibility of the rest of the contributors, the programme, La Sexta, and the journalism profession. Among the posts that related this specific event to journalism as a whole, Figure 3 shows how the manoeuvres of a single journalist could affect the overall "dignity" of the profession. This particular tweet was linked to external content on a similar case of information manipulation that targeted another leading politician, Juan Carlos Monedero (López, 2022). López's article was not the only one to express an opinion about the status of journalism. Anita Botwin, whose tweets were also included in the sample, published a piece in the monthly magazine El Salto. This was part of the messages that made up the sample in which "Ferrerasgate" was mentioned as another example of the current situation of Spanish journalism: The right to information is trampled underfoot by the major media powers in Spain on a daily basis. What Ferreras has done and what the fourth estate is doing, for the most part, is controlling activists and setting traps for them. (Botwin, 2022) Similarly, Àngels Barceló, director of the radio programme Hoy por hoy, also engaged in the posts under study: Political corruption in this country, in addition to being enabled by certain politicians, would only have been possible with the involvement and collusion of  shortcomings, which they then reported. In this case, media professionals were criticised as much for their lack of credibility as for their lack of independence. To a lesser extent, professional protectionism and partial news content also came in for criticism, hence the demand that Antonio García Ferreras abandon journalism on 605 occasions (Figure 4).
Twitter users mostly posted original text messages but sometimes included original audio-visual content, in line with the creative audience (prosumers). This "memetisation" of the object of study, as shown in the following figures (see Figures 5, 6, and 7), helped to increase the virality of "Ferrerasgate" and become part of pop culture. The content of these could be categorised. All of them (45) were used with negative intent. In turn, the purpose of ridiculing (75.5%) the key actors involved was clearly the main category found, well above those images that included elements of criticism with some hateful content (24.5%).

Purpose of the Tweets: Crypto Hate Speech?
Even though audio-visual editing lowered the tension in the Twittersphere, the tone used was aggressive and sometimes infringed on the individual's right to honour. This type of tweet was classified in the "groundless arguments" category. Sketch Engine was used to analyse the discourse of the posts in this category without researcher bias. When conducting a simple thesaurus search for "Ferreras," the results showed the correlation between this term and other nouns such as "chufla" (joke), "mierda" (shit), "asco" (disgust), and "vergüenza" (shame), as evidence that all these words were used in a similar context, coming to function as synonyms (Figure 8).
The in-depth analysis of the discourse generated by Sketch Engine and the search for the main qualifiers used to describe the host of Al Rojo Vivo revealed some negative terms that were also used to attack his physical      appearance: scoundrel (nine); disgusting and dishonest (seven); corrupt, liar (five); arsehole, vile, and trashy (four); revolting, repulsive, mafia-like, fat, pathetic, terrorist, and repugnant (three); cowardly, shameless, despicable, contemptible, outrageous, and delinquent (two).

Conclusions
The emergence of Twitter and other social media platforms has meant that the journalism ecosystem has expanded, thus conveying information to a greater number of people with less effort. However, journalism has also succumbed to the threats inherent to those platforms. With the advent of Web 2.0, the media have found that a new actor has taken away the prominence they used to have when it came to persuading public opinion, the users. The influence of users, also erected as prosumers, is stronger than ever. Not only do they have the ability to select which media are worthy of their audience, but they also have a much greater critical capacity. This, on the one hand, affects the journalists' practices and, on the other, raises the standards of quality that the profession should have. In this context, new alternatives need to be provided that are focused on extending the use of discourse analysis. Due to the qualitative nature of this method, the use of programmes such as ATLAS.ti (Chavez-Montero, 2021) has sometimes been widespread without resolving the shortcomings that are observed when the sample is as broad and heterogeneous as the one selected (N = 2,846). The three variables that underpin this analysis (object, argument, and purpose) have made it possible to examine the most important connotative components of Twitter messages. This was combined with the use of the software-based algorithmic functions in Sketch Engine, which made it possible to circumvent the second problem that is perceived in work of this type: researcher bias (Dosouto, 2009). "Ferrerasgate," a political-media scandal that reverberated around the Twittersphere, is a perfect example of how these users-prosumers commented on an event that had great repercussions. A series of conclusions can be drawn from the analysis of the selected content. Firstly, in terms of the object of the criticism, the discourse was found to be highly personalised. It focused on the presenter and key player in the scandal, Antonio García Ferreras, who was the subject matter of most of the comments. Ferreras was the person who was fingerpointed most often, along with other professionals from the media industry and some politicians. However, a significant number of messages were critical of his programme, the television channel, and the journalism profession in general. This served to identify some of the problems of journalism, as well as the demands of users. They ranged from the demand for adequate information (truthful and objective, prevailing over entertainment)to outright respect for journalism ethics, professionalism, and the diligent use of information sources, among other conditions.
According to the arguments used, problems such as lack of independence, misinformation, and an absence of professionalism revived old debates and revealed a threatened profession, trapped in a systemic crisis (Bazaco et al., 2019;Brown et al., 2016;Cabezuelo & Manfredi, 2019). Not surprisingly, the same threats pinpointed by users had already been evaluated as decisive factors in the recent loss of credibility in the journalism world and its professionals (Sixto-García et al., 2021). It is true that the component of aggressiveness prevented any constructive approach to these problems. But this was simply a consequence of the nature of Twitter as a sounding board and an ideal space for confrontation (Chen et al., 2022). It is therefore important to note that the rhetoric analysed had a totally destructive purpose, where hollow discourse and name-calling became the prevailing features. This, strongly fed by the dynamics of polarisation and ideological quarrels that pervade the digital environment, makes it necessary to categorise this discourse or, at least, to identify it as a phenomenon. Insofar as it does not refer to vulnerable groups, it cannot quite be defined as hate speech, although the purposes it serves, mainly discrediting and disqualifying, bring it very close to this category. So much so that the latest academic contributions have suggested that this aggressive discourse should be referred to as crypto hate speech and encouraged researchers to study its possible consequences. The term "crypto hate speech" was coined by Herrero-Izquierdo et al. (2022) and may be useful when it comes to solving the terminology conundrum that exists in research related to hate and polarisation (Galán-Muñoz, 2020).
Regarding this last point, it is worth highlighting that emotional argumentation prevailed over rational argumentation in the construction of digital messages, which inevitably connects with the growing use of hate and intimidation on the internet as an intentional or unintentional way of putting an end to any kind of debate. It is not surprising, therefore, that this crypto hate speech may continue to evolve through avenues that are as yet unknown, which may accentuate the existing difficulties in having peaceful coexistence in social media. The confirmation of the three initial hypotheses invites further reflection on the challenges that this type of attitude poses for democracy and, by extension, for journalism. In today's context, marked by the hegemony of the internet as a communication channel, the conflict caused by this type of phenomenon seems difficult to solve. Studies, such as this one, establish an initial basis for delimiting intangible content such as hate rhetoric while, at the same time, identifying some of the challenges facing journalism. While social media and the cybersphere have provided an exceptional opportunity for society in terms of how it communicates, it is no less true that the different forms of hate and intolerance continue to have an increasingly greater range of manifestations. Identifying them will require further research.
Pablo Berdón-Prieto is a predoctoral researcher at the University of Valladolid (Uva, Spain). He holds a degree in advertising and public relations and a MA in communication research as a historical-social agent (UVa). He has published several articles in high-impact journals such as Estudios Sobre el Mensaje Periodístico, Obets, Comunicación y Sociedad, Historia y Comunicación Social, or Memoria y Civilización. He has also attended several national and international congresses such as Spanish journalistic society (SEP) or the Latin Society of Social Communication. He is a member of the HISMEDIA research group and is part of the project Politainment in the Face of Media Fragmentation: Disintermediation, Engagement and Polarization (POLDESPOL), as well as the project CARTODIOCOM: Cartography of Hate Speech in Spain From Communication: Sports, Bullfighting and Politics. Jacobo Herrero-Izquierdo is a predoctoral researcher at the University of Valladolid (Uva, Spain). He is a graduate of journalism with an MA in communication research as a historical social agent (UVa). His lines of work focus on the study of the media and its relationship with political power throughout history. He is currently part of the renowned research group HISMEDIA (History and Media) and has published articles in high-impact journals and several book chapters in prestigious publishers. In 2020 he completed a research stay at the Faculty of Political Science, University of Zagreb. In 2021 he made another stay at the Complutense University (Madrid) as a member of the project CARTODIOCOM: Cartography of Hate Speech in Spain From Communication: Sport, Bullfighting and Politics.