Many of us are still looking for explanations for the tsunami of votes given to antisystem and anti-PT (Worker’s Party) candidatures on the first round of the 2018 Brazilian elections. How were these candidatures with so little television time and smaller budgets declared to the Electoral Justice able to overcome consolidated leaderships with wider access to resources?

The majority of attempts to explain this phenomenon address the internet and campaigns on social networks. Some of these explanations have been trying to associate the Brazilian phenomenon to events that took place in other electoral processes, especially in the US and in the European Union. On the other hand, discoveries published by the newspaper Folha de S. Paulo showed the illegal participation of Brazilian businessmen in sending of mass pro-Bolsonaro spams, with the goal to impact voters.

Two characters symbolize the types of explanation about what happened. The first is Steve Bannon, the political strategist for Donal Trump‘s campaign, editor in chief of the extreme-right website Breitbart News, and Cambridge Analytica’s collaborator. The second character is Luciano Hang, Brazilian businessmen, owner of the chain-store Havan, and Jair Bolsonaro’s supporter. How do these explanations relate to each other, or even, what is happening on the Brazilian internet?

Attempts to explain and the tip of an iceberg

An explanation that concerns Bannon has been around longer during the electoral period. The portrait of a smiling Eduardo Bolsonaro besides Bannon in Manhattan, in August of this year, fed the idea that the ascension of the former Brazilian military captain would be a product of the tactics used by Trump with Cambridge Analytica, the company that was the pivot in a scandal, in April, for having used the unauthorized data of millions of Facebook users for the electoral campaign.

These tactics were revealed by a Cambridge Analytica former employee to the press. The company used a Facebook “personality quiz” to collect users’ and their friends’ personal data and, thus, build detailed psychometric profiles about them. Then, this databank was used to inform the campaigns of Cambridge Analytica clients, serving to customize hyperpersonalized ads to each voter. This is what is called “microdirecting”: using paid propaganda tools on the web allows the boosting of that ad for the newsfeed of that person which broadens the possibilities of convincing someone. Amassing a huge quantity of personal information from voters, Steve Bannon’s strategy would have inflated the conservative field by sending the exact message that voters needed to hear and, therefore, behave as Donald Trump’s campaign wanted, helping his election. The manipulation of feelings in favor of electoral strategies and the lack of transparency regarding the use of this personal data ended up submitting Mark Zuckerberg, Facebook’s CEO, to a long public hearing in the North American Congress and leading to Cambridge Analytica’s ending.

Bannon’s meeting with Eduardo Bolsonaro was the catalyst for the narrative that the electoral results in Brazil would be explainable by this technique of “microdirected” advertising. The idea is that it would be possible to pinpoint a centralized motor of the manipulation of Brazilian voters with great foreign interests behind it. A text went viral on social networks affirming that this scheme would explain the behavior of Brazilians during the electoral process, a thesis that has been frequently repeated by other articles, opinion pieces, and videos shared on anti-Bolsonaro groups.

With the discovery by Folha de S. Paulo that companies supporting Bolsonaro’s campaign paid up to R$12 million in contracts for sending WhatsApp messages “en masse”, the scenario seems to begin to show its real complexity. These values were not declared to the Electoral Justice and the telephone databases seemed to have been obtained legally and illegally. One of the characters that appears in this explanation is Luciano Hang, owner of Havan, and named by the newspaper as one of the participants of the scheme. But the injection of illicit resources and the professionalized communication work is just the tip of an iceberg that still has much to be explored.

The infrastructure of network propaganda

If Steve Bannon has indeed offered some type of support to Bolsonaro’s campaign and his allies, we still do not know. It is also quite possible that, within the next few days, more discoveries surface about the economic power and illegalities behind the production and dissemination of content on social networks.

The attempts to explain these events still lack adequate hypotheses and conceptual formulations in order to better understand the structure of Bolsonaro’s campaign on social networks encompassing all available signs. By observing the development of the campaign and the set of indicators, it seems possible to affirm that the campaign happens through an infrastructure of network advertising. This concept uses the diagnosis of US researchers Yochai Benkler, Robert Faris, and Hal Roberts, of Harvard, in their new book “Network Propaganda”, which discusses how online political campaigns have been exploiting network dynamics to increase their capacities. Of course, these dynamics are localized and supported by circumstantial factors.

In Brazil, the tactics of using propaganda in the style of the “Cambridge Analytica scandal” do not seem to have been used, that is, tactics centered on the purchase of advertising space on social networks to send “micro-directed” propaganda. At least for now, there are no indications of any large campaign made through paid advertisement on the internet by candidates of the Bolsonaro family, not on the accountability provided to the Electoral Justice, neither on the Facebook tool for political paid ad transparency (according to what we have described here [in Portuguese]). Still, PLS (Bolsonaro’s party) candidates, which were the major surprise in the legislative elections, did not have significative spendings on paid propaganda (according to the same sources consulted above), and, when they did spend, they did not show any sophistication in the purchase of segmented audiences, like Cambridge Analytica. Paid propaganda was made but directed to larger audiences with broad segmentations (like through states) and simple contents (like virtual leaflets). Neither are pages of Jair Bolsonaro’s supporters (mapped here), which could be boosting its content to hypersegmented audiences for the candidate, spending significative amounts on Facebook.

At the same time, PSL’s presidential candidate has continued to be the most searched candidate on Google and was the subject of most tweets during the electoral period. Candidates for Congress which were linked to him are influencers with huge followings and sharing rates on Twitter and YouTube. Enormous supportive fan pages on Facebook propagate his ideas, show their characters and make campaign contents go viral.

This lack of investment in advertisements on social networks supports the hypothesis that the political force of this wave captained by Bolsonaro’s candidature is the result of a distributed and spread effort, with a coordinated mix of hired effort and others doing spontaneous and voluntary work, visibly more than his adversaries. In this hypothesis, the campaign communications do not reach users through the hypersegmentation of interests and personality data but from an infrastructure that blends new elements of digital communication with old well-known strategies of political propaganda, like lists of telephone numbers.

A symbol of this type of distributed effort and one of the most used forms of communication among Brazilians, WhatsApp holds an important role in this scenario. As it is an app of cryptographed messages, WhatsApp does not offer advertisement space, nor the possibility of directing messages to specific groups — a message is only really disseminated if someone forwards it or if someone breaches the company’s spam block and is able to send en masse messages (here seem to be the efforts of businessmen like Luciano Hang). Yet, the viral aspect of these messages and their conversion to votes depend on many “voluntary mailmen” all over the country, a mass of content forwarders that make the message reach the “final frontier” (that is, family and friends groups and the individual message flow) and there it is raised.

As it is articulated on a network, this propaganda infrastructure gains efficiency when the group administrators are organized, acting as important “knots” of the web by concentrating many connections and improving the flow of messages. It also gains efficiency when militants spend not only their time and personal network on it but great financial resources and more sophisticated tools, as it seems to be the case with Hang.

Thus, in WhatsApp (and possibly other communication platforms), the administration of such infrastructure seems to count on a mixture of paid and voluntary work, along with a decentralized structure.

From the paid side, there are the spendings declared to the Electoral Justice directed to this type of communication: as it has been recently exposed, Jair Bolsonaro paid R$115 thousand to a Rio de Janeiro company to administer and feed with content at least 1500 WhatsApp groups supporting the candidate. On this side, there are also illicit investments, which are much more difficult to assess, in double illegality: these are undeclared resources which come from companies, which, according to the Federal Supreme Court in ADI 4650 from 2015, are prohibited from financing campaigns. The exposed scheme points to payments for communications companies to provide the services of feeding groups but also for some sort of “spamming” — the online version of the old SMS and telephone list calls.

There are still many doubts about the sophistication of the group administration provided by companies since, as Folha revealed, there is no clarity if they only act in small public groups (accessible through links) or also in private groups, which are more difficult to investigate, and there is also the question regarding how these automation or even segmentation tools work for different kinds of groups. However, it is clear that the setting is as efficient as it delivers propaganda messages to a larger amount of people than its opponents. Efficiency also seems to be connected to the messages’ adherence to the political moment and people’s feelings, thus gaining more chances of being forwarded.

Still, by constituting a network (as the Web itself), the production of images, videos, links, and audios that navigate there does not need to be centralized: it is open to another possible army of volunteers and militants who can also attach themselves to that infrastructure in order to distribute propaganda on WhatsApp and other mediums.

An infrastructure built on time and in the Brazilian style

In May 2017, BBC Brasil was already reporting on the existence of a pro-Bolsonaro militant network, encouraged by the candidate himself, who was part of hundreds of supporting groups on WhatsApp and engaged personally on the communication with his “fans”. This and other accounts document the construction of this network propaganda infrastructure since at least 2013. At the time, these groups began to promote the behavior of “soldiers of the myth [as Bolsonaro has been referred to]” which passed on to other social networks like Youtube and Facebook, and receives orders, like posting the words “Bolsonaro 2018” or give negative reviews to a video with criticisms to Bolsonaro.

The political propaganda network has been built occupying each one of the platforms used by Brazilians, respecting their peculiarities. Each of them is a layer to be taken over and that feeds back the other ones. If the organization of the militants and the distribution of messages can be rapidly made through WhatsApp, the political education can be via YouTube. The dispute of narratives can occupy Twitter and the increase of the network aiming friends and the like can be made on Facebook. All of this following the rhythm of each platform and having as highlights the performances of the leaders before the traditional media, just another one of their stages. This network infrastructure in several platforms enables the political campaign without a unique center for spreading content. According to what has been indicated by analysts, the primordial function of the campaign coordination is now one of “validating” the messages that suit them, and not of exclusively creating them.

When, during the campaign, professional work is hired — illicitly or not — in order to feed these networks, it is attached to an infrastructure that has already been built and, with that, it has a gain of scale and efficiency, as content is forwarded, transformed for other platforms and protected from alternative versions. As we have described, this is the peculiarity of the network propaganda infrastructure: the ease in which stakeholders, malicious or not, can attach themselves to any point of the infrastructure and stir its functioning. Indeed, the recent revelations indicate that in each of these layers, illicit practices might have happened; in addition, as today we know more clearly, all of them are supported by the dissemination of fake and divisive news, which are capable of generating commotion and being passed on. The punitive, the moral and anti-corruption crusade, the hate speech and the anti-PT (Worker’s Party) movement have worked well for these purposes.

Lastly, the existence of a great experimentalism is clear — the investment on trial and error, on pushing their luck on each attempt –, a kind of political-technological Brazilian entrepreneurship.

The ability of this network to shift interpretations, manipulate events, and reach an enormous amount of people became clear to Brazil at this point, but it has been built for years. It centrality has even been admitted by Jair Bolsonaro, who recently criticized WhatsApp’s measures to restrict from over 200 to 20 contacts one can forward a content at once (in response to episodes of violence resulting from the dissemination of rumors through the app, especially in India). This has shown that the company’s decision has tightened the bottleneck, limiting the flow of communications on the decentralized propaganda network of his campaign. A recent article from a journalist and two Brazilian researchers for The New York Times even suggested that the sharing number should be reduced to 5, according to what WhatsApp has done in India to fight the sharing of rumors — a proposition which immediately raised the rage of a Twitter segment, with the hashtag “#CensuraPetista” (Censorship by the Worker’s Party), leading the trending topics hours after the text had been published.

Steve Bannon’s photograph alongside Bolsonaro may hide a more complex picture than a mere Brazilian tentacle of Cambridge Analytica or the existence of an all-mighty stakeholder manipulating people for the expansion of Jair Bolsonaro’s political scope. It is possible that foreign money and power are pulling on the knots of the infrastructure of the network propaganda being addressed here. Still, it is also possible that this whole infrastructure has been the product of a genius and Machiavellian planning, with undertones of military strategy, as much as it can be just the addition of disconnected factors.

In any case, it is necessary to recognize that undoing the abuse of economic power and the use of resources and techniques to manipulate the public opinion will have to go through the recognition and the investigation of the operation of this infrastructure; through the development of renewed juridical theses about the accountability of the campaigns for these acts, and of knowledge about the technologies and the techniques being put to use, by the Justice system and the campaigns themselves, which make the denounces; and also through solutions of technological design by companies, who need to think carefully in order to not have them backfire. At last, the important question about why many people you know became a knot inside this network and why this strategy worked is yet to be answered. Then, maybe perspectives from sociology, anthropology, political sciences, economy, and psychology might help us more.

Originally published in El País Brasil on October 19th, 2018

By Francisco Brito Cruz and Mariana Giorgetti Valente

Translation: Ana Luiza Araujo