Waves of disinformation in the Brazilian elections

A coalition of 24 media outlets verified and debunked 147 rumors circulating online during the presidential elections in October

By Sérgio Lüdtke, (Versão em português brasileiro)

Misinformation in the 2018 Brazilian presidential campaign moved in waves and like the tides obeyed the gravitational force of the public agenda. For 12 weeks, I sailed in a sea of ​​misinformation trying to trap rumors and lies that emerged on the web with astonishing speed and strength.

In three months, a team of two editors and I, along with reporters from 24 newsrooms, participated in Comprova, a coalition of media outlets that verified and debunked rumors and lies circulating during the Brazilian presidential campaign. We found these rumors by monitoring social networks in search of problematic content that might damage the electoral process.

During the project, I saw rumors and lies move like waves. After flooding an environment, the same rumors recoiled to return later with force, to reach different spaces and people. The currents that commanded this flow followed the public agenda and the moral issues that dominated online conversations. For example, fake election polling content appeared shortly before and after the real poll. Its creators had an environment where there were low levels of voting information; just enough that people knew that a new poll had taken place but without knowing its results. But when lies circulating among friends and family in closed WhatsApp groups, that lie can be more easily accepted.

We saw increasingly sophisticated content that was created to misinform the public during the project, which took two paths: complexity and simplification.

There was a complex conspiracy theory that reached more than 1.5 million views in just over 24 hours via a viral video. In the video, a man doubts the result of the elections and says that he will announce the results based on a mathematical equation. For six days, journalists from four Comprova partner organizations investigated the video and concluded that it was impossible to determine fraud from the equation presented in the video. The Comprova report was published one day before the date promised by the man in the video for disclosure of ballot fraud and, we hope, reduced the impact of the rumor.

Some rumors used simplified tactics, like lowering the image resolution to make verification more difficult and to contextualize the information differently. One rumor used a video that was recorded two years earlier during demonstrations against President Dilma Rousseff’s impeachment. The video was used to illustrate a claim that a car had been destroyed by Workers Party protesters for having a sticker supporting Jair Bolsonaro. The resolution was reduced and only red flags and a crowd around a white car was identifiable in the video. The video was enough to offer “proof” to support the false narrative. In this case, that seemed to come to a dead end, we were lucky and found the original content in a single Google search.

The other sophisticated element to these rumors is the absence of verifiable information: the creation of lies by suggestibility and no online content. A rumor that Comprova was unable to verify was that “a large weekly magazine” had received 600 million in bitcoins to destroy leading candidate Jair Bolsonaro. By insinuating this rumor, it then made the first magazine to publish an unfavorable report about Bolsonaro considered to be the recipient of the bitcoin. This bitcoin rumor contained no actual content so our team could not prove it to be false.

There were a few dozen other dubious claims checked by Comprova that were never published because it was impossible to reach a definitive conclusion. Even so, the team improved its ability to identify and accurately verify misleading content, which allowed us to publish 147 stories in 12 weeks. Most of the stories published — 135 — were found to be false. In some cases, we managed to stem the spread of rumors; in others, the rumors returned in different formats. Unfortunately, most of the misleading content Comprova debunked remains online.

There are still serious challenges for any team that investigates rumors:

• Respect for privacy and freedom of speech prevents us from finding the DNA of content that travels in apps like WhatsApp, which is capable of creating disinformation tsunamis.
• Finding original content and establishing provenance is the most challenging problem for journalists working on verification.
• In Brazil, candidates and parties create their own verification sections on campaign sites and some rumors debunked by Comprova were “denied” by candidates and parties. These denials eventually gained strength and made us aware that “artificial victimization” may be used in the future in favor of candidates.

Constant monitoring, a more active stance by journalists to intervene when necessary, and the sharing of experiences on projects like Comprova will be the key by which together we can reach a more healthy and controlled information environment.

(Versão em português brasileiro)

Previous articles about Comprova:
- What WhatsApp “API Access” Meant for Comprova
- Comprova wraps in Brazil
- In an age of misinformation, sustainable newsroom collaborations should become a priority