Racism, Death, and Protests in Brazil: digital and traditional news coverage of Black Awareness Day after a racism crime on Twitter

2.1 Protests, press, and social media

Social media platforms were essential in organizing social movements during social unrest and political upheavals. Recent history has given examples of social upheavals managed via social media channels and gained media attention, such as the Arab Spring and Black Lives Matter (Browning 2013; Freelon 2018; Gerbaudo 2015). However, scholars have shown how the news media delegitimize protests by placing a higher value on angles involving conflict and spectacle, which shape norms and routines of news production (Kilgo and Mourão 2021). For example, the press relies more on official narratives from police and elected officials than protesters’ views (McLeod 2007). Thus, the media’s protection paradigm exacerbates social conflict by taking sides. This pattern has persisted in different movements and times (Boyle et al. 2012).

Social media somewhat disrupts the paradigm, as users are not as critical of protesters as journalists (Harlow and Kilgo 2021). This has made social movements rely on social media to become relevant and reach wider audiences (Bennett and Segerberg 2012; Harlow and Harp 2012). The Arab Spring protests were historically marked by considerable attention focused on the role of digital technologies in promoting collective activism to circumvent state-operated media channels. In other words, social media platforms gave voice to these people that were otherwise not represented in the traditional media (Overbey et al. 2013). However, academic literature has shown that the relationship between news media and protests was established in the following manner: “the input was the content characteristics, and the output was public engagement metrics for each specific story” (Kilgo and Mourão 2021: 4). As such, this was not exclusively influenced by users’ intentions but is “also constrained by algorithmic curation that might privilege one type of content over others” (Kilgo and Mourão 2021: 4).

Thus, the news media have embraced the opportunities afforded by online social networks to bring attention and traffic to their news portals (Myllylahti 2020). On the other hand, the affordances of social media have impacted the gatekeeping role of the press, redistributing the power traditionally held by news organizations to users (Bachmann and Gil De Zúñiga 2013; Gil de Zúñiga et al. 2012). In this space, the public has a dynamic performance through comments, sharing, verification, and recommendations (Bruns 2003). To overcome this disadvantage, news media intends to maximize digital news coverage by legitimizing news frames.

News organizations use frames as “a central organizing idea or storyline that provides meaning to an unfolding strip of events” (Gamson and Modigliani 1987: 143). In prioritizing some facts and details over others, journalists foster a particular interpretation of that event, yielding distinctive values in audiences (Chang 2015). For example, Harlow et al. (2017) examined the visual news coverage of protests in news articles that included images or videos. They found that news stories with legitimizing frames were more likely to be shared than those without them. Therefore, the legitimization of protests by traditional media is essential, as these organizations hold power within online social networks (Holbert et al. 2010; Robinson et al. 2017).

In a different vein, digital news organizations promoted a profound reconfiguration in the media industry, changing news products, organizational structures, work practices, and audience representation (Salaverría 2019). During the 2010s, these digital news outlets became a “blooming phenomenon,” leading to several cases of success, such as BuzzFeed and Vox (Bødker 2017; Tandoc and Foo 2018; Tandoc and Jenkins 2017). Generally speaking, digital media perceive their audiences as tech-savvy and multimedia-minded, allowing these media to use innovation to adapt to their public’s context. In this scenario, alternative forms of news media have increased, giving rise to organizations that provide a voice to other publics by adopting radical, counter-hegemonic content and different news values, such as alternative aesthetics and non-advertising models ⁠⁠(Holt et al. 2019).

Thus, alternative media have been primarily perceived as an option or even opposition to the legacy or mainstream media (Buyens and Van Aelst 2021; Harcup 2005). Similarly, alternative media is associated with democratic, open, and non-hierarchical values used by “radical, leftist political agenda and close ties to social movements” ⁠⁠⁠(Holt et al. 2019: 861). On the other hand, there are criticisms of the absence of professionalism, the lack of financial and organizational stability, and living up to their own ideals instead of promoting information to the general public (Downey and Fenton 2003). In contrast, traditional media is characterized by its uniformity and professionalized journalism ⁠⁠⁠(Holt et al. 2019). Despite these opposite views, both types of media are vital for society in many regards. These two kinds of organizations tend to portray protests differently. In this respect, news media relies on frames that depend on the norms and routines of their organization, which shape news production and could directly affect user engagement with protests. While much less is known about how digital media outlets depict protests, traditional media tends to follow a more delegitimizing pattern concerning racial issues than stories covering protests about immigrants’ rights, health, and the environment (Kilgo and Harlow 2019).

In online social media, news coverage significantly impacts people’s thought processes, where information consumption is no longer an independent activity. Instead, it forms part of a continuous connection to the digital space that relies on abstract principles to avoid the loss of connection to the simulation and retains users. In other words, non-click behaviors are common because platforms intentionally want people not to click on the content so that they spend time viewing social media content (Ellison et al. 2020). In this respect, news media coverage has also influenced the users’ perceptions of protests on online social media (Harlow and Kilgo 2021). Shahin et al. (2021) identified that Black Lives Matter (BLM) suffers from hybridity with other movements in different countries after its extensive coverage.

Thus, it is crucial to explore how the media environment shapes political communication about socio-political issues, sometimes in unexpected ways. This study, therefore, contributes to a better understanding of news framing by digital and traditional news outlets and how they influenced the users’ perceptions of racial protests on online social media.

3.1 Framing, protests, and media

Regarding media effects, scholars have shown that news framing can impact audiences’ perceptions of protests (Entman 1993; Shoemaker 1982). Additionally, news media can decrease public support for such events and weaken these social movements (Harlow and Johnson 2011; Harlow and Kilgo 2021). The perception in the framing literature is that frames are particularly influential in changing how “people understand, define and evaluate issues and events” (Carnahan et al. 2019: 1).

Essentially, framing involves the selection and salience of topics. “The central logic of framing is that journalists construct the symbolic representations of society that members of the public use to make sense of events and issues” (Bronstein 2005: 785). In a simple way, frames can promote a particular problem definition, causal interpretation, moral evaluation, or treatment recommendation (Entman 1993). This study builds on prior framing research in this area (Hertog and McLeod 2021), explicitly seeking to identify the frames associated with protests and social movements.

From a general perspective, frames can represent all the knowledge about a particular object or event in news stories. Entman (1993) proposed four framing functions found in a single sentence, although many sentences in a text may perform none. To define the nature of the problem related to an event described, the news media seek to “determine what a causal agent is doing with what costs and benefits, usually measured in terms of common cultural values” by adopting the “problem definition” frame (Entman 1993: 52). In the “causal interpretation” frame, the idea is to “identify the forces creating the problem” (52). The “moral judgment” frame is considered to evaluate causal agents and their effects. Last, the suggested remedies frame “offer and justify treatments for the problems and predict their likely effects” (52). Scholars have reviewed the media framing studies and concluded that Entman’s media frames had remained the most influential in understanding news frames (Matthes 2009).

Using this paradigm, Tonetto Beraldo and Herrera (2021) studied frames in tweets of two newspapers and their public around a controversial political decision that took place in Brazil in 2019, which decided to cut funds for public universities as a way to mitigate the effects of the economic crises that the country was facing. Their findings suggest that traditional media follow the same pattern in portraying this event by adopting the exact words. Furthermore, the causal interpretation was the most used frame, meaning that the tweets displayed the event, focusing more on the search for those responsible for the measure.

However, news media tend to negatively represent protests confronting institutional norms and the presumed status quo (Harlow and Kilgo 2021). This pattern is described in the literature as the “protest paradigm” (Chan and Lee 1988), with news framing as a critical aspect. Recent studies have used news media frames to design a more subtle pattern to the protest paradigm, including legitimizing and delegitimizing perspectives of protests and debates. For example, Kilgo and Mourão (2021) found that the news media’s delegitimization of BLM has resulted in lesser support for activists and more criticism toward protesters. Conversely, other countries have seen movements merge with BLM to legitimize them (Shahin et al. 2021).

In a delegitimizing view, three frames—confrontation, spectacle, and riot—are used to understand the different lenses adopted by news organizations to portray protesters and their actions negatively. The confrontation frame emphasizes conflicts with the status quo and typically between protesters and police, highlighting the violent encounters and number of arrests. When protests are depicted as dramatic or sensational aspects are stressed, it is described as a spectacle frame. On the other hand, protesters are described as destructive or threats to society in the riot frame. In addition, this frame tends to focus on these protesters’ violence and deviant behaviors.

The press portrays these protests under the debate frame in a more positive light, which “puts the social critique brought by the social movement at the centre of the coverage” (Mourão and Kilgo 2021). Because the news coverage on social media about protests is a reasonably recent phenomenon, most framing research has yet to explore how purveyors of information on social media might influence the users through their posts. Considering the preceding literature, this study sheds light on the role of news media on Twitter during the abovementioned case of racist killing in Brazil. In the next section, we describe, in detail, the methods used for the analysis.

4.1 Data collection

To understand the news coverage of Black Awareness Day and protests after a racist crime in Brazil, we rely on a mixed-method approach. An initial list of news media organizations mapped by Atlas da Notícia’s survey (2020) was used to approach this phenomenon. We could identify the Twitter handles of Brazilian news outlets and other relevant data used for our data analysis from this list.

With these registers, we built a Twitter list of news organizations in Brazil and collected data using Twitter’s Application Programming Interface (API). Launched in early 2021, Twitter Academic Research API helps scholars secure millions of public conversation historical and real-time data points, limiting its data collection to 10 million monthly. The platform’s API is quite powerful compared to its competitors (Ahmed 2021). Our dataset includes tweets from news organizations that published content between November 19, 2020 and January 1, 2021. In total, 267,576 tweets were collected.

To better exploit our dataset, we eliminated all tweets which did not contain the terms “Carrefour” (the place of the event) and “João Alberto” (the name of the killing person). This allowed us to have only the news media’s tweets concerning the case. Thus, our final database contained 1,500 tweets and their metadata.

4.2 Data preparation

To better analyze them, we built a computational model responsible for preprocessing our data. The first step was converting all tweets to uppercase, reducing the complexity and error chances of the analysis. Generally speaking, tweets come with much knowledge and can help generate relevant and actionable insights for scholars. However, they also bring some noise to the words. For example, users, including news organizations, rely on hashtags and emojis to make their tweets more attractive and draw the public’s attention. Therefore, it is crucial to eliminate unnecessary stop words and punctuation signs to understand these messages’ context and meaning. Similarly, the signals of “RT” from the retweets and mentions (@) should be removed to leave only the relevant content. In our case, hashtags (#) have been kept as they aggregate information in our tweets.

In addition, we concatenated any entity names, common expressions, or names of cities that were compound. In such cases, composing terms such as “Rio de Janeiro” became RIO_DE_JANEIRO. These steps were relevant to analyzing how the news media portrays the protesters and the killing through words. In the following subsection, we describe the processes behind our data analysis in detail.

4.3 Data analysis

We combined qualitative and quantitative methods in different phases of our data analysis. Using the data mapped by Atlas da Notícia’s survey (2020), we could classify these news outlets according to their region and type (digital or traditional news media). In our study, we considered newspapers, network/cable TV, and radio news as traditional news outlets, while organizations described by Atlas da Notícia as digital were considered digital media. In this context, Atlas da Notícia describes digital media as not only digital-born organizations but also news portals, which began in the print age and transformed into digital businesses or mere news aggregators or news search engines that have become an entrenched presence on the Internet. These data are relevant to Brazil because of the country’s physical structure (Furtado 1984), grouped for administrative purposes into five greater regions—Northern, Northeast, Central-West, Southeast, and Southern—reflected in the Brazilian media ecosystem (de-Lima-Santos et al. 2022). The country has one of the most commercialized media systems, posing risks to pluralism and the representation of different national realities (Guerrero and Márquez-Ramírez 2014).

Similarly, the media market is in the hands of big commercial conglomerates, resulting in a high media concentration. These conditions result from the country’s historical view of commercial and private media as democratic (Ganter and Paulino 2021). Furthermore, these organizations are based in wealthy areas, which are most concentrated in the Rio-São Paulo area.

Our analysis also correlates the collected data with the racial data from the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística, IBGE). As mentioned, Brazil is the second black nation in the world, only behind Nigeria and the first outside Africa. However, Brazil’s systemic racism still reigns in the country (Mileno 2018; Shahin et al. 2021).

In the second step, we used the clean data after the preprocessing phase to identify the most frequent words and associate them with the four frames proposed by Entman (1993): problem definition, causal interpretation, moral evaluation, or treatment recommendation. This allowed us to overview how the different types of media framed the developments after the killing and during Black Awareness Day on Twitter.

In order to study news media frames on Twitter during the protests, we filtered tweets containing the terms protests, protesters, and manifestations in both singular and plural to detect tweets focused on protests and related topics. In total, 70 were identified as associated with these terms. Inspired by the work of Tonetto Beraldo and Herrera (2021), we analyzed these tweets to understand how the news media portrayed protests and protesters. Then, we followed a qualitative analysis to classify this content. Our research combined both the “protest paradigm” (Chan and Lee 1988) and “four framing functions” by Entman (1993) to analyze news media frames. For this, we considered only the seven frames: (i) problem definition, (ii) debate moral, (iii) moral evaluation, (iv) confrontation, (v) causal interpretation, (vi) riot, and (vii) spectacle. In our understanding, the frame “recommendation of solutions” encompasses the same goals of “debate.” Therefore, our analysis includes only the latter. Table 1 describes the frames used to categorize these tweets.

Two coders were responsible for this analysis. Both of them coded every tweet separately. No full news articles were used in the coding process. In case of dispute, they together judged using documented sources, expertise, or other contextual information. We conducted intercoder reliability tests to determine if there was an agreement between the two coders. The results showed a Krippendorff’s α ≥ 0.87 for the coding, which is above the required value (α ≥ 0.800) and much higher than the acceptable rate, as “α ≥ 0.667 is the lowest conceivable limit”. Only a small number (less than 10%) of tweets were in the case of disputes. In those cases, coders had to decide whether or not the relevant party agreed with an opinion.

5.1 The coverage by segment and region

In our tweets, we could identify that Freitas’ killing and its events were more proportionally covered by digital news media than traditional news outlets. In absolute terms, the proportion of tweets from digital media outlets (59.33%) was higher than traditional news organizations (40.67%) based on the 1,500 tweets published. It means there were more tweets from digital news outlets than from traditional media, although the rate of tweets per digital media was 7.8, while in traditional news media, it was 11.8. In other words, the traditional media news outlets proportionally posted more tweets than the digital news outlets, as shown in Table 2. This is in line with previous studies that showed that mainstream media still hold power within online social networks (Holbert et al. 2010; Robinson et al. 2017).

As mentioned before, this event happened in the South region of Brazil. Although the area had great coverage in the average tweets per outlet (6.85), it is behind Southeast, Central-West, and Northeast. As indicated in the Atlas da Notícia’s survey, the Southern part (Southeast and South) concentrates more news organizations than the Northern and Northeast. These two regions have a more significant proportion of news deserts or near deserts in the country. Furthermore, these desert areas are equally distributed all over these two territories. In the Southern region, specifically in the most prominent Brazilian cities, such as São Paulo, Rio de Janeiro, and Belo Horizonte, the online segment acquired predominance; about two-thirds of the news organizations are digital (Atlas da Notícias 2020). However, our data indicate that the Northeast and Central-West have proportionally given more attention to the case than the Southern region, as shown in Table 3.

To understand if this has a racial correlation, we looked at data from IBGE. As shown in Table 4, the Southeast and South regions are composed of majority-white people. Although Brazil is very mixed compared to other countries, the Northern, Northeast, and Central-West regions are formed of people who consider themselves Black or Mixed. Prior study has indicated that Whites and Asians are equally resistant to Black Lives Matter’s ideas, highlighting the interethnic tensions and hostilities (Kilgo and Mourão 2019). A similar situation might have happened in the news coverage of these regions. As media reproduces human behaviors, our findings have indicated more interest in reporting the case online in Northern, even with the higher proportion of news deserts (Atlas da Notícia 2020). Without considering the Southeast (where most newsrooms are concentrated), the Northeast published 266, followed by the South (233), where the event occurred.

In terms of which segment covered the case, our results indicate that tweets in the Southeast and North regions came mainly from digital media outlets (see Table 5). This is an interesting finding from the North perspective. As the Northern part has higher barriers to adopting digital technologies due to rainforests and less investment (compared to other regions), radio is traditionally the most common medium in these states (Atlas da Notícias 2020). Our results show that this violent act was more salient in the agenda of digital news outlets in the region than the traditional media.

5.2 Framing the killing case

Of the 1,500 tweets, we analyzed the most frequent words to understand how the news media portrayed the case (Tonetto Beraldo and Herrera 2021). In our analysis, we could identify that the most common terms used in the tweets of Brazilian news organizations were in their majority based on two news frames: “problem definition” and “recommendation of solutions” (Entman 1993). For example, terms describing problem definition, such as the local of killing (CARREFOUR, SUPERMERCADO, PORTO_ALEGRE, RS, BRASIL), the actors (JOAO_ALBERTO, BETO_FREITAS, SEGURANÇA, SEGURANÇAS), and the acts (MORTE, MORTO), were commonly found in tweets of digital and traditional media outlets. Similarly, the frame recommendations of solutions could be identified in terms, such as justice (JUSTIÇA), police actions (POLICIA), manifestation (ATO and PROTESTO), protesters (MANIFESTANTES), and Black Awareness Day (CONSCIENCIA_NEGRA). There were also many similarities in how traditional and digital media framed the tweets.

One term caught our attention despite the commonalities between digital and traditional media. Many digital news outlets used the hashtag #Blacklivesmatter (#VIDASNEGRAIMPORTAM). Contrary to the findings of Kilgo and Mourão (2021), who found a delegitimization of Black Lives Matter by the news media, our results suggest that digital media support the Brazilian activists and protesters. Similarly, the impact on the American culture of Black Lives Matter can be seen beyond the United States. In this context, BLM resonates with local concerns, as described in the literature as a hybridization of social movements (Shahin et al. 2021). This aligns with the perception that alternative media has closer ties to social movements ⁠⁠⁠(Holt et al. 2019). On the other hand, there is less evidence of this support by the traditional media, as this term did not appear in the most often-used words. Table 6 shows in detail the news frames that appeared in the 30 popular words of each news organization.

5.3 News media framing protests

When it comes to protests, scholars have provided evidence to show that news framing can delegitimize and denigrate social movements (Entman 1993; Shoemaker 1982). Having this information, we qualitatively analyzed the tweets dedicated to protesters and terms associated with protests.

Combining the “protest paradigm” (Chan and Lee 1988) and “four framing functions” by Entman (1993), we have identified that, although both media have focused on defining the problem, there are significant differences between the way digital and traditional media portray the protesters. A brief description of frames used by each media is given in Table 7.

In the traditional media, negative aspects were highlighted or adopted a blasé attitude by associating the protesters with Black Awareness Day or cases of internal attention, such as George Floyd (Barrie 2020). For example, a tweet from Correio do Povo, a newspaper from Bahia (Northeast), posted a tweet with a delegitimizing perspective about the protests: “Carrefour dawns with closed doors and remnants of protests” (Carrefour amanhece com portas fechadas e resquícios dos protestos, in Portuguese). Another example was Exame, a business magazine from São Paulo, which posted a tweet negatively portraying the protesters: “Carrefour: stores crash in protests and global CEO speaks out” (Carrefour: lojas são quebradas em protestos e CEO global se pronuncia, in Portuguese).

We could recognize frames of riot (25%) and moral evaluation (15%) aside from problem definition (55%). Few tweets opened up for the debate (5%). None of these tweets provided a causal interpretation or posed a confrontational view of the protesters. Most tweets only presented the descriptions of facts or the problem, for example, a man killed in Carrefour.

On the other hand, tweets from digital news media gave salience to debate (20%) and more causal interpretation (26%) in addition to problem definition (44%). This is partly explained by Brazilian news outlets covering forgotten and marginalized communities, such as Favela em Pauta (de-Lima-Santos and Mesquita 2021; Paiva 2018). For example, the alternative media Mídia NINJA called for a protest in Carrefour using the hashtags #Blacklivesmatter (#VidasNegrasImportam) and #BlackNovember (#NovembroNegro) (Tweet in Portuguese: "Vidas Negras Importam: Manifestação amanhã às 18h no Carrefour! #VidasNegrasImportam #NovembroNegro).