Low impact of bots and fake data, but at least one “fake” article among the top stories

The first presidential debate ahead of Brazil’s October election took place on Thursday, August 9. An analysis of online engagement around the televised event concluded that automated profiles comprised only around 3 percent of total interactions and that statements classified as “fake” by fact-checking agency Lupa (a partner of FGV DAPP’s Digital Democracy Room) had little impact on broader conversations. Yet for the first time since the pre-campaign period began, one piece of unverifiable news appeared among the top ten trending articles on social media. The piece, published by the website República de Curitiba, claimed Twitter “removed” mentions of right-wing candidate Jair Bolsonaro from the platform during the debate.

Bolsonaro, one of thirteen contenders for the presidency, is second in the polls behind Former President Luiz Inácio Lula da Silva. With Lula’s candidacy in question as he serves a prison sentence for corruption charges, Bolsonaro–a former army captain with a strong social media presence–is expected to take the lead.

Mapping Interactions

The above interaction map illustrates the proliferation of social media discussions around the debate. From a collection of 1.070 million tweets and 643,000 retweets comprising the candidates’ names, six clusters stand out. The central node of the network (35.2 percent of the profiles participating on the discussion) comprises jokes and ironic comments made about the candidates’ performance — these were primarily targeted at Bolsonaro and the evangelical candidate Cabo Daciolo. The blue node (19.2 percent of total profiles), represents those in favor of Bolsonaro, while left-leaning profiles and those supporting Former President Lula (8.8 percent of total profiles) appear in red.

The more right-leaning purple group (13.1 percent of total profiles) comprise anti-Bolsonaro accounts, supporters of former São Paulo state governor Geraldo Alckmin, and those calling for the participation of center-right contender João Amoedo in the debate. In light pink (17 percent) and orange (3.6 percent) are left-leaning groups not interacting directly with Workers’ Party (PT) supporters. The light pink group opposes Bolsonaro but does not show support for any specific candidate; the orange group is composed primarily of supporters of left-wing candidate Ciro Gomes, who generated mainly positive references in the debate.

Analyzing Automated Profiles

Only around three percent of interactions were influenced by automated profiles, or bots. This percentage can be described as low when compared to recent discussions around the Brazilian political debate. For instance, when FGV analyzed the level of bot influence in the July 2018 debate, it discovered up to 22 percent of interactions were motivated by bots. Both analyses were made using the same methodology, developed by FGV DAPP and available here.

“Fake” Information Mentioned by Candidates

Fact-checking agency Lupa classified eight of the candidates’ references during the debate as “fake.” When analyzing the impact of this information, FGV DAPP found that two topics Lupa investigated mobilized users on Twitter: the gun control debate, connected to Bolsonaro, and printed voting, related to Cabo Daciolo. In general, however, users discussed the themes without citing the wrong data mentioned by the candidates. Ciro Gomes’ wrongful claim that the wife of judge Sergio Moro — one of the leading names in the Car Wash probe — received housing allowance, was an exception. Gomes corrected the information during the debate, and his apology Tweet was among those that garnered higher levels of engagement.

Two examples show how discussions played out around fake information cited by the candidates. Online discussions around Rede candidate Marina Silva’s public sanitation platform contained more complaints about how she discussed the topic than references to the wrong data she cited on number of households with inadequate access to infrastructure. Former Central Bank president Henrique Meirelles’ statement about Venezuelan migrants received more engagement as a result of his position against closing the borders.

Social media users also discussed Cabo Daciolo’s comments asserting that Ciro Gomes integrated international left-wing groups such as Foro de São Paulo and Ursal (Latin American Union of Socialist Republics — a fictitious group). These were mentioned 80,400 times. Among the tweets that circulated the most was humorist Danilo Gentili’s criticism of the potentially controversial behavior of the candidate (4,300 thousand retweets). While jokes and ironies also spread, a few people took Daciolo’s comment seriously. Profiles that mentioned Ursal engaged in some 6,000 interactions.

Volume of mentions on Twitter regarding information considered “false” by Agência Lupa

Fake News

For the first time since the beginning of the pre-campaign FGV DAPP’s analyses identified among the top-stories related to the debate an article labeled as “fake” by fact-checking agency Lupa, according to the number of engagements on Facebook and Twitter (between Thursday 12:00 p.m. and Friday at 12:00 p.m.). The piece published by the República de Curitiba website claimed Twitter had removed mentions to Bolsonaro on the platform. The claim that Twitter had removed hashtags related to the candidate, which made them disappear from the Trending Topic list, echoed on the social platform, though not originating from the mentioned link. More than 32,000 references were made to the subject. The most retweeted publication was from Bolsonaro’s profile, with more than 7,700 retweets.

1.6 Million Twitter Mentions in Three Hours

In all, the TV debate generated 1.59 million Twitter mentions about the candidates between Thursday at 10:00 p.m. and Friday at 1:00 a.m. The hashtag #DebateBand entered the world’s Trending Topics list. The most popular hashtags were #debateband (572,000 mentions), #bolsonaronaband (209,000) and #estoucombolsonaro (177,000). The most engaging topics, besides irony, jokes and comments about candidates’ attitudes, were themes such as corruption, economics, and security.

This article was originally published by Fundação Getúlio Vargas’s Diretoria de Análise de Políticas Públicas (FGV-DAPP). FGV-DAPP is a partner on #ElectionWatch Latin America, a joint initiative between the Atlantic Council’s @DFRLab and Adrienne Arsht Latin America Center.

To learn more about FGV DAPP’s methodology for bot identification, see here.