Mitigating Election Disinformation in Brazil

Audio of this conversation is available via your favorite podcast service.

It is well understood that for all the shortcomings of the tech platforms’ approach to elections in the United States, it’s much worse abroad, where often language and cultural barriers combine with fewer political and business incentives for firms such as Meta, Twitter, YouTube and TikTok to properly resource elections.

Now, just weeks before a general election in Brazil that will decide that country’s next President, there are signs that disinformation is rife on the platforms, with many observers concerned about the potential for violence. To learn more, I spoke to two experts involved in efforts to identify and mitigate election mis- and disinformation in Brazil:

• João Brant, coordinator of desinformante, an initiative of the nonprofit Ponteio Comunicação, Information and Culture and the Instituto Cultura e Democracia in Brazil, and
• Flora Rebello Arduini, Campaigns Director at SumOfUs, a global activist community that seeks to curb the growing power of corporations.

In the past month, both organizations published reports on the challenge in Brazil:

• Read The Role of Digital Platforms in Protecting Electoral Integrity in the 2022 Brazilian Election from desinformante
• Read Stop the Steal 2.0: How Meta is Subverting Brazilian Democracy from SumOfUs

What follows is a lightly edited transcript of the discussion.

Justin Hendrix:

I'm so pleased that the two of you can join me today and help Tech Policy Press listeners understand the situation in Brazil and understand what you're up against in what is now just weeks until the general election there. But, first, perhaps we can start there with a bit of what is the situation.

I understand that the polls, while they're not necessarily tight between the two top contenders, they are tightening up. Of course, there's a major problem yet with election disinformation. Would one of you be willing to just give a sort of general sense of the situation at the moment?

João Brant:

Brazilian elections will run on October 2nd the first round and October 30th the second round. It's a very peculiar election because we have Bolsonaro, who is the incumbent, running against Lula, the former president that was the most popular president that left government with 88% of approval, but then was put in jail. Then the 22 cases against him were dismissed. So everyone knows quite well the candidates, and so that creates, I would say, a different situation than previous elections where disinformation had more space to flourish. But still, and yet we see very serious cases of disinformation running on this election. And it's definitely a concern.

Flora Rebello Arduini:

If I may add to what João just said. I think precisely because we know the candidates very well, and we know that elections in Brazil since the last 2018 presidential election, we've seen this polarization. And so polarization is like a great situation for disinformation, misinformation, hate attacks. So, you can see this offline kind of polarization in Brazilian society, surely being reflected in the online space as well. So, you're seeing these two figures and their supporters also reflecting this kind of content online.

Justin Hendrix:

So, both of your organizations have just released reports kind of on the problem of mis and disinformation in this election cycle in Brazil. I'd love if you will, to just sort have you maybe explain some of your top line findings, the way you characterize the problem after your study of it.

Flora Rebello Arduini:

To start with SumOfUs research, we were really trying to understand how Facebook in particular, Facebook, Instagram, so the Meta products, which are incredibly popular in Brazil. Brazil is one of their main markets worldwide. For example, Brazil was the second market of WhatsApp only after India. So, we were interested to see how matter was putting in place its policies ahead of the elections in Brazil. And September seven was our focus to research because September seven in Brazil, it's a national holiday, it's the independence day. And since Bolsonaro came to power, it became a political date, right? It used to be just a holiday, a national holiday, not a political holiday, but since then, since he came into power, it became a far right and right holiday in the country. And we were interested to see how this politicization, I'm sorry if that's how you say it, was being reflected online and was being reflected on Facebook.

And so wefocused one week of research and what we found was, basically we divided in three pillars. So, the first pillar was electoral campaign ahead of its sign. So, it was actually illegal electoral campaigns. In Brazil, you were only allowed to run electoral campaigns after August 16th. And we assessed several ads. So, impersonated paid ads from the far right candidates on Facebook. So, the second pillar were ads that were straightforward anti-democratic ads. So, they were pulling speeches from Bolsonaro, just calling people to go to the streets on September 7th. And the reason why they're so concerning is because they're really calling people to go fight for their lives on September 7th, which is a very similar narrative that we saw when Trump was calling US citizens to go to Capitol Hill on January the 6th.

And the third pillar was what we found on encrypted platforms. So, Telegram and WhatsApp. And again, there, we found even more aggressive rhetorics, obviously because there is no moderations in this platforms, WhatsApp and Telegram. Then the speech is way more violent and way more aggressive. Obviously people learn how to navigate the platform's policy. So, they know that if they use a specific words, it would be easier for the AI to catch and to block certain ads or certain posts. There is not such a concern on WhatsApp. And so you could see these three kind of streams of content being spread on this platforms. And I think it says a lot. One of the findings is that, how much has Meta learned since January the 6th in DC? And our conclusion is that it learned pretty much nothing.

So, why we say this is, I know it might sound just headlining and just flashy kind of formation. But the real reason is we have no idea what Meta is doing ahead of the Brazilian elections. If you look at their website, the policies that they put in place and what they say they put in place ahead of the Brazilian elections are extremely generic and you can see that they use global numbers. They don't say, for example, how many Brazilian, Portuguese moderators there are working ahead of the Brazilian elections. All of these things show that... Anyways, they're putting again the profits before elections, integrity and democracy. So, for us, our biggest concern and which is aligned with several other organizations that João, and desinformante is also part, is that again, there is absolutely no transparency from Meta about exactly how they're putting in place policies and actions that are indeed tailored to the Brazilian context and tailored to the elections in one specific country.

Justin Hendrix:

Now, Facebook said in a statement that's in the press, of course, that it has launched tools that promote reliable information and label election related posts that it's established a direct channel for the superior electoral court to send potentially harmful content for review. And that it is closely collaborating with Brazilian authorities and researchers. João, you are a representative of a coalition of researchers and groups, perhaps in speaking to some of the findings in your recent report, I'd ask you to address that. Do you regard Meta or the other social media platforms as effectively working with civil society in Brazil?

João Brant:

Well, I think the big issue is they definitely are not doing enough. When they say, when they mention these kind of measures that they adopt, that they are really adopting that. And it's fine. And I think it helps some uninformed citizens to understand where they can vote or whatever that's useful, yes, definitely it is. But we're talking about a really big problem that it's not being addressed.

I will separate into three main concerns. One is, we looked at the policy, we look at the policy of all the companies, the different platforms here in Brazil. And I can say that none of them is really enough to combat disinformation considering not only that this information against the election or the electoral system, which has been a big issue here, but also the disinformation against, or from one candidate to the other, which is one of the big issues in any of the elections.

So that's one problem. It's the policy, but of course they're different. If you take, for instance, Twitter, I would say that it's probably the strongest policy that we have for protecting the electoral integrity itself, although they don't really do any kind of assessment on disinformation from one candidate to the other. If you take Meta policy, that's the only company that doesn't consider disinformation. For instance, the allegation of fraud without any kind of evidence or elements. And so if you can allege that elections are being frauded and you are not being considered someone that is promoting disinformation, and that's only Meta that has adopted this loose perspective. And so there's different perspectives on policy.

The second thing is I would say the enforcement. And so we had global witness doing an experiment showing that Meta, hasn't been enforced their own policy to protect ads, for instance, against basic information, the election, and that created a reaction, Meta did change that policy to not accept ads that go against the legitimacy or that contest the legitimacy of the Brazil elections. But if you see how it's going in place, it's probably hasn't changed in anything there, some studies to be published in the next weeks.

Okay. And the third thing is related to the examples. So, the way they reapply that policy. YouTube for instance, it says that any conservation theory that can lead to violence or anything, it's forbidden in the platform. But the example is a QAN and issue that for us has nothing to do with our reality. And we do have conspiracy theories that are serious, that can affect violence in real place. So, I would say the big picture. What we think that it's really could be done is to protect the platforms against the idea that they will be the place to arrange civil unrests and against the democratic order in Brazil. I would say that that's our main concern on 20 days before the election, which is if you are offering place that people can organize and really go promote an approval against the democracy in our country, it's a big problem. And apart from Twitter, all the platforms don't have policy that prevented this to happen.

Justin Hendrix:

I do want to just pause on this and get both of your perspectives. I mean, clearly coming from the United States and we have this tendency to, of course, be very US-centric in our point of view, and I'm sure I'm not in any way immune to that. Of course, I've spent a lot of time talking about and researching and writing about January 6th in the United States and the role of social media in that phenomenon. How concerned are you at the moment that this election cycle may result in violence? When I look at your reports, many of the images, the memes, the presence of weapons, direct incitements to violence look very familiar to me, as someone who has looked at the 2020 cycle very closely here. I don't know which one of you'd like to take that.

Flora Rebello Arduini:

I think just connecting a dot before just what João was saying around the platform's policies, Meta has a specific policy around election ads, right? And social topics. And when we did research, when we ran the research, the ads, the most problematic ones, we selected from a sample of 3000, then we focused on 16 that were specific around elections and social topics. And Meta's policy says that they will remove any content that can lead to offline violence. And all of these ads that we saw, those 16, just 16 got 650,000 impressions. And those were paid ads so, there were, one, monetizing on violent speech, and second of all, they were violating their own policies, as João was mentioning before. So you're seeing that they have the policy. They haven't actually deployed and enforced the policy.

And to answer your specific question around if there is concern about the violence in Brazil, surely everyone is concerned that this is a possibility and you see authorities in Brazil putting in place safety measures that have never been put in place before in Brazilian history. Specifically, if we look at September seven, which was a very key moment for our democracy, there's been security and safety measures, historic ones. There was an actual fear that what they call lone wolves will actually go and invade buildings and institutions. But I don't think it ends at the election. I think as we have to learn of what happened in January the 6th, the rhetoric and the speech will go right way after the election results. And also I think it's important that we are seeing candidates change their discourse, focusing on what is most valuable to gather more supporters and get more votes.

So, we saw Bolsonaro having a very violent speech for years. Surely, his campaign understood if he continue with such violence discourse of military coupe, violence on the streets will probably not gather more votes. So, you could see a little bit of a mild, a bit more mild rhetoric ahead of this September seven. He did some pretty concerning statements in the manifestations on September seven. So, I'm just saying this because we will see a fluctuation of the candidates discuss specifically the far right is specifically Bolsonaro is yet to see how far he will keep going with the attacks against institutional authorities in Brazil. And of course, depending on the results, then we'll have again, to be really careful of where that can take supporters.

João Brant:

The elections in Brazil are really quick in showing the results. So in this first round, results will be clear in the night of the election because it's an electronic voting system. And if Lula wins, probably Bolsonaro's supporters will go to the streets and try to trigger civil unrest against the election result. If it's on the second round, it may happen as well, and so it's kind of unpredictable what can happen and it will happen quite fast. It will be a spark that is put on the streets, it's sent to the streets, and there are lots of military and paramilitary groups with guns and weapons waiting for something like this to happen, or for them to be provoked to go to the streets and start something.

I may sounds really bizarre or pessimistic, but it's sad. It's clearly sad and they're being outspoken about that. So, definitely if Lula wins, we can wait for the results to be contested by the Bolsonaro supporters because they actually won't accept the election result. And that's the moment that the platform will have to change immediately the way to deal with that, or they have to be prepared to understand how can you really not feed an anti-democratic movement, that will probably happen a new be in place.

Justin Hendrix:

So, even in this intense moment, I'm interested in whether João, especially you, thinking about the coalition, if you can, I guess, look at the situation from a historical perspective, is this the way that it is now? Is this the way elections in Brazil will be in the social media age? Will this type of coalition always be necessary? This type of research, Flora that you're doing always be necessary? Can we imagine getting to a better place, or do you suspect that the types of, I suppose, civil society emergency response that you're engaged in now will have to follow every cycle?

João Brant:

Well, unfortunately I think we will have to follow every cycle and that's related to the way the platforms really change at the public debates and the information environment in the last 10 years, that's related to the segmentation, the fragmentation the promotion of extremist discourses. And it's basically related to the business model so, it is a structural debate. Of course, being a structural debate doesn't mean that you have only to discuss structural responses to that because otherwise you keep doing only the big debate on regulation or whatever, but we have to do both at the same time. And so we have to discuss more structural responses and we have to act upon the mediated facts that it has, not only on elections, but I would say on social rights as well. If you take the racial justice movement, if you take the environmental, which is related in Brazil, the climate crisis that we're living, everything is being affected by these informational environment.

Justin Hendrix:

Let me ask another question, I guess, related to this bigger picture, which is that we've seen the types of coalitions that you're working with come together in other countries. Of course, there've been recent ones around the French election. There was a group that I've had onto this podcast in the past, of course there was the election integrity partnership in the United States. And on some level, these groups have themselves, occasionally become part of the narrative, become part of the conspiracy narrative among some populous groups on the right who seek to de-legitimize election outcomes. And I guess in the States, we've seen even the discourse around the problem of mis disinformation become politicized, even though the problem is very much an asymmetric one, that becomes problem. How do you contend with that, both of you in your organizations? How do you both recognize the reality of what's happening and also somehow maintain a nonpartisan perspective?

Flora Rebello Arduini:

So, SumOfUs, it's not based in Brazil. So we have an institutional organization in legal responsibility of being completely nonpartisan in the Brazilian election specifically. So, for the Brazilian research, for the Brazilian efforts that we are having ahead of this election, it's completely nonpartisan. We just look at the facts and we look at the data. So, for this research, for example, we didn't just focus on when we were doing the research on Facebook ads library, we weren't necessarily looking at, "Oh, well, let's see what the far right is saying or what the left is saying." We use neutral terms. We use more lefty kind of hashtags or far right common hashtags.

And it's a fact that the far right pushes this narrative more and on social media, it's just a fact. If we look at the body of research though, not just from SumOfUs, but other great organizations out there throughout the years, you see the far right pushing this narrative way more than other political spectrums.

And so just to say, we will always assess if there is problems and there is disinformation or misinformation being pushed, no matter by what candidate or what political spectrum, this has to be called out. So, SumOfUs it's practically fully funded by members. And so just regular people, just regular citizens. So, it's almost as if it's in our DNA, it's our responsibility to serve people. And so we have people that support different candidates and very varied and across the road because it's a global organization. And so I guess the way we will make sure that we are always looking at it from a macro perspective is just seeing what the data is offering us, which is a problem as João and you Justin knows that we have no transparency from those platforms.

And so finding the data and conducting this kind of research is really hard and it's getting increasingly hard. If you try to look specifically at Meta, if you specifically, Instagram and Facebook, so Meta products, there is one specific tool that is called Crowd Tangle, which is brilliant. It's one of the best tools you can have as a researcher to have a glimpse, but a better glimpse of what's actually happening in those platforms. And they're increasingly cutting off the access to this tool. First of all, Meta bought the tool. So, there is already a concern there. They bought the tool and now they are the gatekeeper of who has access to this tool.

And so for the biggest efforts that we do in researchers in Brazil or elsewhere, is that we don't have one, the lack of transparency from these platforms. We've been saying this over and over again. Second, the biggest issues that because they're not transparent, they don't say, for example, how much of research they're putting in specific languages. They always focus on English and we know the biggest researchers are put in English. And the third is the lack of access. And so I guess one of the policies and one of the focus of our organization is legislation, to make sure that we have strong bodies of enforcement and regulation across big tech industry. We really campaign a lot in Europe for the passing of the Digital Services Act, so the DSA. It's a landmark legislation. So, we want to make sure as much as possible that the good bits of this legislation, which there are many are also passed through across the road. And also making sure that the US, which is their homeland does pass strong legislations that will keep these platforms on their toes.

João Brant:

Can I add something on that? I think Brazil has something that goes over the top of the problems that the global north already knows, which is WhatsApp and Telegram being used as a viral and really mass communication tools. That's Brazil, that's India, that's Nigeria, that's part of Southeast Asia. It's like populous countries that for some, I would say, anthropological reasons and cultural reasons, use a lot group systems and the viral features that these tools offer. And that creates a huge part of the public debate to be buried, to be under earth, to be really not seen by the public. So you can't respond to anything that is happening there, you can't actually understand, you can't see what's happening. And WhatsApp is the most popular tool of information in Brazil today, so that creates a really a concern here.

And of course they adopted some friction measures, something to kind of prevent viralization, but they are really important and things go viral and things go without any kind of accountability and any kind of liability be that legal or moral because part of the issue would be better if we had moral accountability or moral liability for what you say. But if you're not subject to that, if the non anonymity is taken as a general rule and not as an exception, that creates a big problem for the public debate.

Justin Hendrix:

And perhaps just in closing, I want to kind of come back to regulatory reforms you'd like to see in Brazil specifically. But João, could I just push you on this question of, with regard to the function of these coalitions? As a model that's evolving now around the world in democracies, how do you keep them from becoming part of the broader debate about the legitimacy of the information ecosystem? How do they avoid becoming, or is it just sort, I guess, par for the course, that of course, anyone protecting democratic debate and democratic institutions would necessarily be pulled into the orbit of conspiracy theories?

João Brant:

I think part of that, you really can't avoid, it's something that will happen. And of course, the ones that are being at the forefront of these debates will be attacked, will be part of conspiracy theories and et cetera. But the second thing I would say that in our view, the main strategy was to define the document as our common place. We are not to fix coalition. There are fixed coalitions that deal with digital rights, for instance, in the country. These coalitions are signing this document. So, the coalition on digital rights, the coalition on Black movements, the coalition on indigenous movements, coalitions of NGO, they are all signing these documents. And so they are given support to a specific approach to this debate and that's our agreement. It's a kind of document that it's itself our agreement. And at least it covers the fact that no one is speaking different things on behalf of others, where what is poll and defined is on the document.

And so we work careful on the document to really protect from more biased perspectives. We all have our political views, our political perspectives, and that's fine. And I think we are seeing, as Flora said, that the extreme right is yielding more disinformation to go against democracy. And part of that is because there is a kind of affinity between the affordances of these tools and a strategy for populist right leaders to act upon. So once they found this out, they really started using that and that's fine. If the platforms offer you something like this, they will use it. There is no way not to use it. And so I think one thing is to recognize that it's a different perspective, but the other thing is to avoid that this kind of coalitions, or be there more long term or short term are taken for specific proposals. And so far, I think we're doing that. We achieved that.

Justin Hendrix:

So, I'll put this last question to you both. And really, it's a question about the regulatory environment in Brazil and perhaps after this election, what types of reforms you can imagine pursuing that might address this issue going forward? Flora, you've spoken about the Digital Services Act and in some of its measures, I'm sure, you'd like to see replicated. And you've, of course, mentioned the importance of the United States getting its act together with regard to regulating these companies. Unfortunately, I don't necessarily see that happening in the near term. What do you hope will happen there?

Flora Rebello Arduini:

It's a complex issue. When I talk about regulation on big tech and on social media, I usually do a parallel with environmental legislations or conventions, right? Because they are always evolving and they can be just so fixed and so steady in the present that it will freeze any action from legislators about new issues that might come up with the technology. So, it has to be bodies of legislations that are flexible enough to promptly respond to the developments of the technology, but also strong enough that are doable and practical. So, it is complicated. It's a hard kind of legislation and it's new, we've never done this before, to regulate this kind of industries that are so rapidly evolving and has so much power. They're huge monopolies. Meta is a huge monopoly, Google, Apple, so on and so forth. So, just thought it will be helpful to give this context a little bit.

What I think, it's nickname as the fake news bill in Brazil, and because it started from right after the presidential elections in 2018, where we saw the infodemic on disinformation on social media. And so it has this spirit of to fight off disinformation. There are several problematics with this bill currently, as it is right now. And one of them is, for example, parliamentary immunity. If you spread disinformation and if you use public money to post dis misinformation online, so that has to go. But other elements that I really think that DSA could and should have done better, but there is the intention and there is the DNA on DSA that I think would be great to see in legislations across the road.

Specifically, I think the profiling. So, when we're talking about surveillance advertising, and we mentioned this briefly before, is their business model. These platforms, they live from profiling. So, they are able to collect an unprecedented detail, data and information from you gather this information and then sell back to you. So, you're getting this loop, endless loop on social media, that it goes way beyond social media. It goes on online imprint. So, surely legislation and regulations around profiling of how much data these platforms are allowed to take from users.

Also, how this data is utilized. We are seeing mismanagement of personal data on a daily basis, on different levels. From misusing our data, for example, WhatsApp tried to pass a policy last year across the road, where if you had to accept, in the global self specifically, you had to accept that your data would be taken from you and sell to Facebook, or you would have to leave the platform. So, it was a false choice between... Brazilians, nine out of 10 of you use this platform. If you don't accept our abusive terms, you're kicked off of the platform. So, it was just a false choice, for example. So, these kind of data usage has to be addressed.

Surely I hope that more access to researchers and academics and independent parties to assess how the platform is deploying and implementing its policy, how is actually responding to ongoing legislations across the country specifically, and also making sure that they give enough access so that researchers can really have concrete data, concrete assessment of what's working, what's not working and what needs to be done. So, giving this access to independent researchers, I really believe it's crucial, so auditors basically. And also again, so we don't get too late in the debate, it's AI. We have to look at AI as like we're late, it's overdue. So, AI is something that it's popping across the road. We're seeing EU passing an AI legislation, we're seeing Brazil already addressing this topic. So, surely something that we need to pass as soon as possible, but also responsibly.

Justin Hendrix:

And maybe I'll tweak the question slightly for you as well, which is just to say yes, and all of those tech reforms, but in the face of political elites that espouse lies, do you reckon that those types of reforms will ever get us close to some reasonable solution to these problems?

João Brant:

I think at least in Latin America, this is very clear and that affects for instance, why Meta does protect the political discourse, knowing that politicians are the biggest spreaders of this information. I know, and I assume, and I have some empathy with the argument that they are local leaders, they were elected and they have their discourse discussed and scrutinized, I think that's valid for only for the big leaders. And even though you assume something like this, you have to understand that politicians, especially in Latin America, are the biggest spreaders of this information.

We had in Brazil, a parliamentary inquiry committee on the pandemic. And it has a 200 pages chapter showing how disinformation worked. And the conclusion is quite clear that the politicians were the key diffusers, the spreaders of disinformation, right? So, that's one issue that's related to something that the platforms could do themselves. The second thing for me is on regulation, the big thing. I would say that in the last 10 years, we really let aside all the modern values that use it to organize our informational environment, since the postwar in 2010. So, we had 60 years where the professional journalism was the organizer of the public sphere, if we can say like this, an organizer of the democratic regulation systems in Europe, for instance. So, you had all the discussion on pluralism, on diversity and reliability of information was not a big issue because professional journalism would protect and actually did protect that from being a big problem.

So, once you had the social platforms to assuming this new role, it really changes the picture. And we are in a kind of digital state of nature, a Hobbesian state of nature in a digital world, because you really let aside all the social contract that you had with about the public sphere, and you just say, "Go. Go and do your stuff." And your worst psychological perspectives will lead the way we organize information. So, if our regulation doesn't go, at some point, to this core issue, it won't sort out anything. And so for me, it's very difficult of course, to say that because it goes onto the business model of the companies and you can't just put the toothpaste back to the tube, but at some point you have to do that. And I think UK is trying to do that on the online safety bill. I think Europe with this risk assessment that is stated and provided by the DSA, kind of experiment that, and I would say that we have to go in some way for this direction.

Justin Hendrix:

Each election cycle in these democracies seems like a live experiment and I wish you the best as the experiment plays out in your country. And I hope that perhaps we'll be able to discuss how all the variables fit together after the fact, and perhaps arrive at a better place in the near term, so I wish you both the best.

João Brant:

Thank you. Thank you very much, Justin, for the opportunity. And I hope you're right. I hope we have peace in this and monotony. That's what we're expecting as the best scenario.

Flora Rebello Arduini:

Yes. Thank you very much, Justin. It's been a pleasure.