How bot accounts on Twitter are pushing hashtags pro Bolsonaro

Category: Young journalist

Country/area: Brazil

Organisation: Agência Pública

Organisation size: Small

Cover letter:

Dear Madam or Sir,

I am very happy to submit my application for The 2020 Sigma Data Journalism Awards. I am a brazilian journalist, who just graduated at University of São Paulo, class of 2015. Being considered for this award is a great honor and would mean a big step for my career. 

Journalism has brought me a lot of opportunities and adventures. In my first year of college, I took part in a project where I would research, write and record short sum ups of international broadcast news for the university radio along with other colleagues. 

Also during my graduation, I was a reporter at Jornal do Campus, a journal produced by journalism students that covered news around the university. This was where I started with data journalism. Along with other 2 colleagues, we did a fact-checking piece on affirmations from the university dean at the time.

Outside of college, my first paying-job was at an educommunication NGO, called Viração, where I worked at a news agency and a magazine ran by young people and children. The organization purpose was to empower young people through communication techniques, such as journalist, photography, and media literacy. They had a project with Google where we would do activities to teach children about cybersecurity. Through this project I had the opportunity to go to Google Campus in California and talk to other young leaders about social networks, privacy and fake news. This was when I started to get interested on this subject. 

I’ve put together my interest in online behaviours and data journalism by working with fact-checking during brazilian elections in 2018, at the Truco project, from Agência Pública. After six months covering a campaign filled with fake news pushed by social networks algorithms, I decided to investigate more in depth the problem of desinformation online. Still in an internship at Agencia Publica, I was mentored by experienced journalists such as Natalia Viana, one of the few journalists who were with Wikileaks at the Cablegate leak in England in October-November 2010.

I learned how to use TCAT to scrap data from Twitter, have made a course and attended an event from the brazilian investigative journalism organization (Abraji), and used this experiences and tools to produce pieces on how debate is being manipulated online. I use data to investigate this issues and journalism to bring it to light. 

In august 2019, during the fires that took place in the Amazon Forest, I investigated a couple of hashtags on Twitter that accused NGOs of starting the fires and supported the brazilian environment minister, Ricardo Salles. By analysing more than 100 thousand tweets, I found automated accounts that engaged in the hashtags and wrote a piece on it. After it was published, I received a lot of attacks and threats online. But that didn’t stop me. A few months later I found that one of those automated profiles was still pushing pro-government hashtags on Twitter and made another piece on it.

It has been a long journey to get here and I don’t mean to stop it so soon. Winning this award would be a confirmation that I am in the right path.

Thanks for the consideration

Description of portfolio:

For the last year working at Agencia Publica – first as an intern and then as a reporter –, a brazilian organization of investigative journalism, I have been looking into the desinformation environment online. 

One of the most important pieces I wrote about it was the one I described above. In august 2019, during the fires that took place in the Amazon Forest, I investigated a couple of hashtags on Twitter that accused NGOs of starting the fires and supported the brazilian environment minister, Ricardo Salles – #AmazoniaSemONG and #SomosTodosRicardoSalles, meaning “Amazon without NGOs” and “We are all Ricardo Salles” in English. With the help from researchers from Atlantic Council, I got access to the first 50 thousand tweets from each hashtag. Both of them reached the Trending Topics in Brazil for two consecutive days.

From the data, I could track the first tweets with the tags and the users who used them the most. With that, I managed to analyse and plot into graphics how many times the hashtags were tweeted per hour. 

I also found that only 1% of the users who used the tags were responsible for more than 20% of the tweets with the tags. The profiles who used the hashtags the most had several evidences of automation, such as high volume of tweets (more than 400 per day), frequent use of tags, text disconnected to the tags, links to other websites. Among them there were 5 profiles that were very similar. They retweeted each other, had the same profile pictures and names – evidences of what is called a botnet.

Besides the bots, I also found the tweets that were most retweeted with the tags. Adding all this elements up, the piece made a good mapping of how a network of bots was connected to influential pro-government accounts and how they were manipulating the public debate by rising hashtags to Trending Topics. To discuss the issue, I interviewed researchers that look into the desinformation environment on social networks, specially Twitter. 

I also asked the platform about their policies against automated accounts that try to manipulate the TT metrics.

When published, the piece had more than 11 thousand shares in Publica’s website and more than 2 thousand reactions on Twitter and 1,9 shares and likes on Facebook. It was also republished by other 26 news websites, including El País. 

After the publication, some of the accounts charged as robots were excluded. Others remained.

So I made another investigation on one of those that were still active. By using Twitter’s TCAT, I scrapped data from 1 month of activity of the bot account. The result was a dataset of almost 10 thousand tweets, which I analysed by frequency and content. I found the account has tweeted 368 per day during november of 2019 and I found a pattern in its tweets – most of them contained pro-government hashtags and links to a conservative Youtube channel. According to specialists, this is a strategy used by many accounts and it would be clarifying to expose that to public.

The piece also showed which tags were used the most and profiles that the account has mentioned. It was shared by more than 2,7 thousand people on Publica’s website and republished in other 20 news websites.

The pieces are both in portuguese and linked on this application.

Project links: