A Desinformação nas Crises e a Tragédia no RS

O Rio Grande do Sul vive uma tragédia sem precedentes, em parte decorrente das mudanças climáticas e, em parte, decorrente das ações e omissões de governos e pessoas na legislação ambiental (como por exemplo, a permissão de construções praticamente dentro de áreas de inundação e rios).  Essa situação é dramática e sublinha a importância da comunicação em situações de crise e, particularmente, da gestão desses processos em plataformas de mídia social. Foi necessário, em muitos casos, fazer com que informações e pedidos de evacuação chegassem com urgência para certas parcelas da população em algumas localidades. Em outras, foi preciso pedir atenção. Ainda em outras, concentraram-se esforços em encontrar pessoas presas em suas casas e solicitar socorro.  Porém, toda essa situação também sublinhou a falta de planejamento e preparo para o fenômeno da desinformação na crise.  Faço aqui um breve texto com considerações sobre essa questão.

A desinformação nessa crise parece atuar em dois níveis. O primeiro, mais contextual, trouxe conteúdos falsos e enganosos que buscaram tentar desacreditar e culpabilizar autoridades políticas, seja pela imputação de ações, quanto omissões que implicam que essas autoridades estavam diretamente atuando para piorar a tragédia (vide exemplos a seguir, retirados de checagens).

Post Aos Fatos

Figura 1: Imagem do Twitter.

Na figura acima (Figura 1) vemos uma checagem da agência Aos Fatos apontando que a Fazenda estadual não estava barrando veículos com auxílios para as vítimas das enchentes.

Post Lupa

Figura 2: Imagem do Twitter.

Na figura acima (Figura 2), a agência Lupa desmente que o presidente Lula tenha patrocinado o show de Madonna no Rio e não tenha enviado recursos aos atingidos no RS.

Essa desinformação, quase sempre de natureza política, atinge e percepção coletiva da tragédia, instrumentalizando-a para atingir autoridades políticas e ações políticas. 

Também vimos um segundo tipo de desinformação, mais pontual e específica. Esta acompanhou notícias e fatos em localidades específicas e circulou de modo a gerar alarme, horror e mais pânico junto a população. Esse tipo de conteúdo, por exemplo, imitou conteúdo oficial indicando áreas de risco (porém com áreas diferentes marcadas), trouxe pânico ao afirmar em regiões não atingidas pelas cheias que haveria desabastecimento de comida, relatos falsos de corpos boiando e etc.  Quase sempre com tons conspiratórios (“saiba o que ninguém quer que você saiba”) , esses conteúdos minam a confiança nas autoridades públicas, confundem a população e pioram a situação.

A figura a seguir, (Figura 3) outro exemplo, desmente a informação de que teríamos 300 corpos encontrados em uma das cidades atingidas.

Figura 3: Imagem do Twitter.

Na imagem a seguir (Figura 4), um jornal explicita que não havia “desabastecimento generalizado” de gasolina, o que tinha levado muitos a uma corrida aos postos para abastecer (e gerando, assim, pela altíssima demanda, desabastecimento).

Figura 4: Imagem do Twitter.

Esses dois tipos de desinformação, contextual e específica, geram tipos diferentes de efeitos na população. No primeiro caso, a desinformação gera descrédito da população nas autoridades (principalmente governamentais, mas também, científicas, por exemplo). No segundo,  dificultam a chegada de informação relevante e urgente, amplificam os problemas e afetam, diretamente, a tomada de decisões ruins pelas pessoas. Enquanto o primeiro caso foca o confirmation bias, o famoso “viés de confirmação”, extremizando posições políticas, o segundo foca o pânico, o medo, e o sentimento de desespero.

Como esses conteúdos problemáticos circulam com força e rapidez pelas plataformas mais fechada de mídia social, como os mensageiros (WhatsApp e Telegram, por exemplo) é mais difícil que sejam rapidamente desmentidos, pois demoram até chegar às autoridades. Eles se aproveitam das mesmas vias que as pessoas utilizam para se informar em ocasiões como esta, gerando visibilidade e credibilidade para esses conteúdos, pois circulam dentro de grupos e canais onde há o objetivo de informar-se e obter conteúdos relevantes. Eles monetizam a tragédia.  E as consequência são gravíssimas.

Obviamente que a desinformação na crise é  um problema de governança das plataformas. Embora a desinformação tenha existido sempre, são essas infraestruturas que lhe conferem capacidade de monetizar, ganhar audiência e visibilidade.  Porém, é também um problema que precisa ser enfrentado na comunicação de crise. É preciso, assim, também pensar a comunicação de crise com a desinformação nas plataformas em vista.

E o  que podemos fazer para combater a desinformação nesses momentos? Faço algumas sugestões:

  • Ter canais oficiais de conteúdo institucional nessas plataformas, e usar esses canais de modo a não permitir que se tenha um vácuo informacional que possa ser preenchido por desinformação.
  • Monitorar para atuar rapidamente no espalhamento de desinformação ESPECIFICA, que pode trazer consequências piores no curto prazo. Esse tipo de desinformação precisa ser prioritária em crises como a nossa, pois gera caos imediato e piora a situação de modo geral, inclusive, colocando as pessoas em risco.
  • Atuar de no combate sistêmico à desinformação CONTEXTUAL, principalmente através do apoio a um ecossistema de combate à desinformação, com a atuação de agências de checagem e veículos jornalísticos. Embora esse tipo de desinformação seja igualmente importante, ela não tem efeitos imediatos, mas é o seu acúmulo que gera descrédito.

 

 

REPORT: Public Discourse, Democracy and the Brazilian Presidential Election on Meta’s Facebook

Overview of results: 

  • The discussion and views about democracy in Brazil are very polarized on Facebook’s public groups and posts. This polarization is based on two groups: pro-Bolsonaro and pro-Lula.
  • Each group uses the same concepts but with different connections, indicating a dispute over the meaning of these discourses;
  • Pro-Bolsonaro’s cluster uses democracy connected to the defense of a “coup” since they argue there will be fraud in the election;
  • Disinformation and conspiracies about the Supreme Court, the electronic ballot, polls, media, and the election circulate to support this “fraud” discourse on the pro-Bolsonaro cluster;
  • Democracy in the pro-Lula cluster is connected to the defense of the democratic institutions such as the vote and the ballot and the fight against Bolsonaro’s supporters’ views on a coup.

Context

On Sunday, October 2nd, Brazil will vote for president. Most polls indicate leftist Lula da Silva ahead of the dispute, with the far-right candidate and current president, Jair Bolsonaro as the runner-up, but far from Lula.

This election is probably one of the most polarized ones in a long time in the country, thus, it is expected that the campaign was also very polarized. In fact, most of the campaign was marked by Bolsonaro’s frequent attacks and threats to the election system and the electronic ballot and election fraud claims. The current Brazilian president used several occasions to share his beliefs in conspiracies to take him out and for “communism” to take over the country. Bolsonaros’ speeches during office often used conspiracy theories and disinformation to attack his opposers and Brazilian’s democratic institutions He also often showed support for a military coup and claimed there was never a dictatorship in the country, even though Brazil endured a 30-year one.

Bolsonaro’s populist discourse mimics other far-right populist leaders, based on conservative’s views on costumes (even often the president himself often makes sexual innuendos in public speeches), support of the military dictatorship, criticism of democratic institutions, anti-elites and anti-scientific views, support of religious extremism and others.

The Brazilian Supreme Court and the Supreme Electoral Court both held several meetings to show the electronic ballot process (which has been used in the country for more than 25 years) and allow for security checks, as well as public scrutiny forms to check for election results. In August, several Brazilian authorities, scholars, personalities, and politicians launched the “pro-democracy letter” against Bolsonaro’s anti-democratic views.

Lula da Silva, on the other hand, rarely talked about the election process and mostly criticized Bolsonaro’s anti-democratic claims and threats. Lula was the president for two mandates, at the beginning of 2000, and is the representative of the Worker’s Party, associated with the left in the country.

In this context, the key question we look into in this research is: How does the concept of democracy appear for different candidates’ supporters in this polarized context? We believe that the idea of conspiracy and coups in the country will be connected to different views of what democracy is – and how the election represents this process.

Methods

To explore this question, we collected data from CrowdTangle, focusing on Brazilian Portuguese and the word “democracy” from the beginning of the presidential campaign (August 15th) to September 22nd. We collected 15076 public pages and groups that posted content about democracy and 36227 links connected to these posts.

To explore the network of links and pages we used Social Network Analysis, focusing on a bi-partite network, with nodes that are links and public pages/groups. We used indegree to map the links that were shared the most and outdegree to show the pages/groups with more activity on sharing these links. We explored the network structure using clusterization algorithms on Gephi.

Based on the network structure, we extracted the most shared links for each cluster and the posts that were used to share them. We further analyzed the type of link shared and the content of the post, to understand how these contents described democracy. For this step, we used content analysis to classify the links and co-occurrence of concepts to create networks of association to describe the discourse from each group.

Polarized link-sharing Structure

First of all, as expected, the network is very polarized, with two big groups of nodes that share similar content and almost no connections between these groups. These networks comprise mostly a cluster of pages/groups of Lula’s supporters (on the left) and Bolsonaro’s supporters (on the right). The clusterized structure implies that both groups shared different links, and each group of links circulated the most within the group (Figure 1).

Polarized structure
Figure 1: Public pages and groups that shared similar content.

The pages and groups are mostly from supporters of each candidate, however, the sources of the most shared links come from different types of sources.

Lula’s supporters group shared mostly links from media outlets (such as Folha, UOL, Metropoles, Globo, and others), as well as the “pro-democracy letter“. From the 30 most shared links, we found 20 links from media outlets, and 10 from other Facebook pages – Lula’s supporters).

On the Bolsonaro’s supporter’s side, on the other hand, most shared content came from other pages/groups on Facebook and youtube videos. These contents often made innuendos about how there was a conspiracy to elect Lula da Silva; criticized polls, saying they were “false”; criticized the Supreme Court and the Electoral Court and the ministers and other institutions. Thus, most of the content shared within pro-Bolsonaro’s cluster was based on disinformation, false images, edited content to make false claims, and conspiracy theories (see example on Figure 2). The most interesting part, however, was that part of the most shared links come from official chambers of deputy and senate candidate pages, Bolsonaro’s supporters. From the 30 most shared links, we found 15 videos from other Facebook pages with disinformation; 5 youtube videos with disinformation, and 10 links from hyperpartisan outlets.

Figure 2: Example of disinformation on the most shared links by Bolsonaro’s supporters. The image compares Bolsonaro’s rallies after the beginning with early pictures of Lula’s rallies, when supporters were still arriving. The image insinuates that polls are tainted.

Democracy as a Conceptual Dispute

Finally, we also focused on how each group described and connected different ideas to the concept of democracy. In the pro-Bolsonaro cluster, the connections to democracy were similar to Bolsonaro’s discourse. In the group, democracy is strongly connected to “freedom”, as “individual freedom”. “Elections” is strongly connected to “cup”, “left”, “communism”, “socialism”, “dictatorship”, “crime” and “journalists”. Other frequent associations are “God” and “Family” (Figure 3).

Figure 3: Pro Bolsonaro’s cluster discourse about democracy.

These connections show that Bolsonaro’s conservative discourse reverberates on these groups. They  hared content that supported their views on how the left represents a “risk” for democracy, as it represents “dictatorship”, “communism” and “socialism”. The general discourse on this cluster is also strongly based on nationalism and patriotism, using words connected to Brazil, Brazilians, Nation, Green and yellow (Brazilian flag colors), Emojis of the Brazilian flag, patriots, and other concepts.

Democracy, for this group, is very connected to the idea of individual freedom and the combat of the ghost of communism and socialism. Discourses connecting disinformation to the defense of the democracy were also common. For instance, some posts argued that the Brazilian “constitution” allows for a “military coup” to defend democracy against the “communists”.  Conspiracy theories connecting the institutions such as the Electoral Supreme Court, the Supreme Court , election fraud and others to a “leftist coup” were also frequent.

Democracy from pro Lula cluster
Figure 5: Concepts connected to democracy from the pro-Lula cluster

From the pro-Lula cluster, we have a very different picture. The concept of democracy is also very connected to “Brazil” and “Brazilians”. However, it is also connected to “defense”, “respect” and “elections”. The “pro-democracy letter” has a central role in the discussion as well. “Vote” also has a central impact in this cluster. There is also the concept of “coup”, but more connected to “Bolsonaro”. This group also has different concepts from the first one, such as “women”, “death”, “hunger” and “work”, which are connected to frequent critics of Bolsonaro’s government (Figure 5).

Several concepts appear in both clusters, such as “Brazil”, “Brazilians”, “rights”, “president” and “power”. In the pro-Bolsonaro cluster, these concepts are connected to nationalism. On the pro-Lula cluster, on the other hand, these same concepts are connected to an anti-Bolsonaro discourse, with critics of the president. This presence indicates that both clusters appropriate similar concepts to connect to their discourse, as a discursive dispute emerges on the meaning and connections of these words. For one cluster, democracy is connected to respecting the election, and voting, with the liberty to choose and vote for president. For the other cluster, democracy is connected to the fear of socialism, a leftist coup, to crime, and fraud of elections. These are two very different constructions around the concept.

Conclusions

This study shows some context on the social media conversation about the election a few days before the date. It shows democracy is connected to different conceptions and the role disinformation has on these different meanings. One of the key ideas is how the election has been connected to a possible “fraud” or “coup” as conspiracy theories and disinformation freely circulate and are legitimated by these networks particularly on pro-Bolsonaro supporters. This phenomenon may create instability and distrust in the process, as voters may see any problem as a confirmation of fraud and particularly, may not accept or believe the possible unfavorable outcome for Bolsonaro.

 

This op-ed is part of the project “Global Democracy Frontliners: Transnational Research Coalition for Tech Accountability and Democratic Innovations Centering Communities in the Margins”, coordinated by Jonathan Corpus Ong with support from Luminate/Resonate. 

It is also supported by the National Council for Scientific and Technological Development (CNPq).

New paper on Disinformation and Covid-19

Our research lab has another paper out! “Bolsonaro and the Far Right: How Disinformation About COVID-19 Circulates on Facebook in Brazil” has just been published in the International Journal of Communication. This paper brings some of the results from the research project about Online Disinformation and the Covid-19 pandemic in Brazil. While much has been said about the influence of politicians and the far-right on disinformation, this is one of the first results that brings some data the reinforces this connection is real and political polarization seem to have a huge influence in driving this type of content. Not only we show that disinformation is drive by political content, but also, that fact-checking is unable to circulate among the same pages and groups disinformation circulates.

Here is the abstract:
This article tackles the circulation of disinformation and compares it to fact-checking links about COVID-19 on Facebook in Brazil. Through a mixed-methods approach, we use disinformation and fact-checking links provided by the International Fact-Checking Network/Poynter, which we looked for in CrowdTangle. Using this data set, we explore (1) which types of public groups/pages spread disinformation and fact-checking content on Facebook; (2) the role of political ideology in this process; and (3) the network dynamics of how disinformation and fact-checking circulate on Facebook. Our results show that disinformation tends to circulate more on political pages/groups aligned with the far-right and Brazilian President Jair Bolsonaro, on religious and conspiracy theory pages/groups, and alternative (hyperpartisan) media. On the other hand, fact-checking circulates more on leftists’ pages/groups. This implicates that the discussion about COVID-19 in Brazil is influenced by a structure of asymmetric polarization, as disinformation spread is fueled by radicalized far-right groups.