Project description

From August, 16th to October, 28th Lupa\’s team worked hard covering what was being said during the presidential elections in Brazil. It fact-checked over 850 statements made by 64 politicians and followed every TV debate with strong live-fact-checking initiatives. Besides that, Lupa has also debunked hundreds of junk information that was being shared on social media and, especially, on WhatsApp. The message app has impressive numbers in Brazil – 127 million active users – and, since it is an encrypted platform, it doesn\’t allow fact-checkers to see what is trending on the system in order to choose what topics should be verified. Willing to work with this \”black hole of information\”, Lupa teamed with Universidade Federal de Minas Gerais (UFMG) and Universidade de São Paulo (USP). Their researchers had developed a system to monitor what was being said and shared in about 350 public WhatsApp groups. And Lupa\’s work consisted of analyzing the quality of that content. In the first round of election, 18,000 people shared on those groups more than 850,000 contents (texts, images, audio files and videos). So Lupa\’s team decided to focus on images – especially on those that had been most shared. The results were terrifying. Lupa studied top 50 images and found out that only 4 of them were totally true. This was the seed for a strong OpEd co-written by Lupa\’s founder, Cristina Tardáguila, and researchers Pablo Ortellado (USP) and Fabricio Benevenutto (UFMG). This article was published in English and in Spanish by the printed and online versions of The New York Times and consisted of a call to action for WhatsApp. The authors suggested the app would make three moves – in order to reduce the spread of misinformation. One of them is now active in Brazil. Besides handling the impact of finding a way to verifying popular content shared on WhatsApp, Lupa also worked in two other directions. It kept producing content to national media outlets, such as Folha de S.Paulo newspaper, Metrópoles website and also Yahoo! and Terra portals. And kept its news literacy program, LupaEducação. In 2018, Lupa\’s team trained three Regional Electoral Courts in Brazil – in the states of Rondônia, Mato Grosso and Pernambuco. In its 4h fact-checking workshops, Lupa has gathered over 100 officials responsible for auditing the electoral systems and the voting process.

What makes this project innovative?

WhatsApp is an end-to-end encrypted platform. It is not possible for fact-checkers (or anyone else in the world) to know what kind of mis/disinformation is trending on the system. So Universidade Federal de Minas Gerais (UFMG) developed "Monitor do WhatsApp" (WhatsApp Monitor, in English) and opened this platform to Lupa in order to help Brazilian fact-checkers see what was being mostly shared in about 350 public groups, where about 18,000 people where talking. This is something that had never been done in the fact-checking world - and probably neither in the journalism community. "Monitor do WhatsApp" is super innovative. It consists of a friendly and easy-to-use platform where Lupa's team could search content extracted from public WhatsApp groups and see how popular it was. The monitor divides the posts in four categories: text, images, video and audio files. It allows searches by dates. So, by using it, fact-checkers can follow a false content that pops up in the monitor and confirm if it is really growing - and should be debunked. The fact that Lupa decided to work with images also brought a sort of revelation to the country. Brazil was used to think about false news as something in the text format. Revealing it could happen by images was important too. Another revelation was the fact that among the 50 most shared images analyzed by Lupa there was no photos of Jair Bolsonaro or Fernando Haddad - number one and number two in presidential polls. This brought up a new side of the polarized narrative during the election.

What was the impact of your project? How did you measure it?

The work done by Lupa, UFMG and USP made international headlines. It was the seed for a strong Op-Ed published by the printed and online versions of The New York Times. The impact of this article was huge. It was reproduced in many languages and made WhatsApp take actions a few months later. Now, in Brazil, it is not possible to forward a message to more than 5 contacts - one of the three suggested actions proposed by the authors. Some important discussion forums were created after Lupa's work became public. The Superior Electoral Court (TSE), for example, put together a meeting for fact-checking platforms a few days later and supported the creation of a collaborative verification project to be held during the weekend of the electoral second turn. Thanks to Lupa's effort, on Oct 27th and 28th, all active fact-checking platforms in Brazil got together and built a 48h coalition to verify false content on social media and WhatsApp. The group managed to detect and debunk 50 false information. Due to some of the content published by Lupa during the presidential campaign, candidates ended up changing the way they talked about certain topics, adopting correct data. This shows the impact of Lupa's "false" and "exaggerated" ratings and the efficiency of fact-checking in the election run. At least three former presidential candidates corrected themselves after Lupa pointed out their mistakes. Jair Bolsonaro (PSL), Henrique Meirelles (MDB) and Guilherme Boulos (PSOL) adjusted numbers they were misusing on TV debates and interviews after receiving negative ratings from Lupa's team. This was even part of an article published by the Poynter Institute. On the LupaEducação branch, the biggest achievement was seen in Rondonia. As a direct result of Lupa's work, a man was arrested after an attempt to fraud the electronic voting machine by applying super glue on it. In the morning of the second round - October 28th - the phone rang at Agência Lupa. It was a member of the Electoral Court in Rondônia, a former student, wanting to review the use of a specific tool Lupa had taught him. The fact-checkers helped him geolocate a publication and the man behind the fraud attempt was caught. Super glue in Brazil costs about 3 dollars. A voting machine, about 600 dollars. In November 2018, for all this work, Cristina Tardáguila, Lupa's founder received ElPeriodico/Grupo Zeta Journalism Award, in Madrid, Spain. She was chosen journalist of the year.
In 2018, Lupa decided to live fact-check all televised presidential debates. In Brazil, those events usually last 2hours and involve no less than 6 politicians and a lot of pressure from their press offices. But since it was founded, Lupa is known by its live tweeting and couldn't keep its fact-checkers from doing it. They just love it! Too much adrenaline. As a straight result of 10 live-tweeting sessions, Lupa is proud to say it has grown in users and also in interactions, without spending a dollar. It only worked organically.
Lupa published a lot more on Twitter - 154% more than its average - and saw the number of followers grow 24%. In consequence, it had an increase of 280% on views and 371% in interactions. Influencers loved Lupa's content. Among new followers Lupa saw not only journalists but also artists, athletes and - of course - politicians. The former general attorney of Brazil, for example, is still one of Lupa's followers.

Source and methodology

Lupa has used WhatsApp Monitor, built by UFMG. Once Lupa's team decided what to fact-check, it would apply its certified method. Journalists first read what has been published about a topic, then find public databases about it. During 2018 presidential election, Lupa has used files from Datasus (database from the Ministry of Health), PISA, Sinopses Estatísticas from INEP (statistics summaries from the National Institute of Studies and Educational Researches), database from Tesouro Nacional (National Treasure), Infopen and Geo Presídios (Penitentiary System), Central Bank, Comex (from the Foreign Trade Bureau), CagEd (from the Ministry of Labour), Anuário Brasileiro da Segurança Pública and Atlas da violência (public security databases from states), and also databases from the World Bank and IBGE (the Brazilian Institute of Geography and Statistics). When debunking images, Lupa's team would work with Google Images, Bing, TinEye, Yandex and InVid. All the team is connected through a Slack channel and a strong Google Drive system. It is important to say that before the campaign even started Lupa's team also requested several information through Lei de Acesso à Informação, Brazilian Freedom of Information Act (FOIA) law. All the answers received through this process where well organized and kept in Lupa's system in order to accelerate the process of fact-checking live.

Technologies Used

Lupa used some tools that were only available to fact-checkers to cover 2018 election. 1) WhatsApp Monitor is a system developed by UFMG. It monitors public groups in the message app, identifying the most shared texts, videos, photos and audios. It doesn’t allow the identification of recipient and sender but helps fact-checkers choose what should be verified. 2) ClaimCheck is a platform that was developed by Facebook and for fact-checkers that are members of the Third Party Fact-checking Project, Facebook's project to fight misinformation. ClaimCheck allows fact-checkers to see posts reported as false by users and how popular they are in the platform. Lupa used it a lot during the election. 3) CrowdTangle is a system developed by Facebook and used by media outlets and fact-checkers. It tracks the spread of content through the internet. It measures the social performance of posts and indicates what has turned into “news”; It also displays the influencers of a certain issue. 4) BuzzSumo is a system which shows the links that are getting viral in different platforms.

Project members

Chico Marés, Clara Becker, Cristina Tardáguila, Douglas Silveira, Plínio Pereira Lopes, Flávia Campuzano, Leandro Resende, Natália Leal, Nathália Afonso, Pauline Mendel, Caroline Lima, Raphael Kapa.


Additional links


Click Follow to keep up with the evolution of this project:
you will receive a notification anytime the project leader updates the project page.