Article: Bots on Social Networks

"The growth of the concerted action of robots represents, therefore, a real threat to public debate, representing risks, in the limit, to democracy," says DAPP director Marco Aurelio Ruediger


Asleep since the demonstrations at the time of the impeachment of president Fernando Collor de Mello (1992), the streets once again became a part of Brazilian politics when the increase in bus fares caused a wave of protests in Brazil in June, 2013.  In that moment, the strategy of avoiding the appropriation of the movement by the political parties – the main targets of the revolt in the streets – and the distrust in the coverage of the demonstrations by the traditional press transformed the social networks into a recognized space for expressing the democracy, since they became protagonists as places for organizing and spreading information.

From 2013 to today, not only were the streets occupied by partisan movements but also the networks were flooded by old political strategies of slander and manipulation of public debates. However, these strategies now happen in a place that allows for the rapid massification of discourses in such a way that puts at risk the credibility of the space and of the information that circulates in it. The traditional pamphleteering for parties, for example, occupies the same timeline of news spread by the press, and so do rumors and detractions propagated by political actors from the whole spectrum of political parties.

The internet and the social networks have become a very important, growing, and dynamic field of the public debate and of the dispute between narratives, which lead to the pursuit of hegemonies in politics. This reality makes room for legitimate and factual discussions, but also for ill-intentioned, illegitimate and nonfactual discourses (fake news).

In addition to this fertile environment for the dissemination of opinions, the automation of publishing tools allowed for the appearance and propagation of bots – accounts controlled by software posing as human beings, which have already dominated part of the life on social networks and actively participate in discussions during political moments of great repercussion.

The study carried out by FGV/DAPP shows that this type of account was responsible for more than 10% of the interactions on Twitter during the presidential elections of 2014. During protests for the impeachment, these bot-provoked interactions represented more than 20% of the debate between supporters of Dilma Rousseff, who made significant use of this type of mechanism. Another example we analyzed shows that almost 20% of the interactions in the debate between users in favor of Aécio Neves during the second round of the 2014 elections was motivated by bots.

In political discussions, bots have been used for the whole spectrum of political parties not only to obtain followers, but also to conduct attacks against the opposition and forge artificial discussions. They manipulate debates, create and disseminate fake news and influence the public opinion by posting and replicating messages in a large scale. For instance, they commonly promote hashtags that gain prominence with the massification of automated posts in order to stifle spontaneous debates on a certain topic.

When we identify bots operating for one side, however, we do not mean to say that the political and public actors situated on that side are directly responsible for the bots in their favor. Several interest groups could be using this type of resource for the dissemination of information. In truth, in a broad sense, there are even bots operating abroad. This actually entices a reflection not only about internal manipulation, but also beyond the national political boundaries, suggesting the hypothesis of the existence of even more actors, strangers to the national scenario, operating these mechanisms on the networks.

The increase in the concentrated action of bots represents, then, a real threat for the public debate, representing risks for the democracy by manipulating the process of consensus building in the public sphere and in the selection of representatives and government agendas that can define the future of the country.

Therefore, identifying these bots becomes a challenge of great importance, since their operation is increasingly refined and capable of replicating human patterns more precisely. Distinguishing the real side with the manipulated one, in the analysis of ongoing social and political processes, is decisive both for the government – whose decision-making process must be anchored in qualified information – and for the civil society, which reverberates the agenda produced on the networks in debates and actions outside of them.

That is why FGV/DAPP developed a refined system that generates content algorithmically for identifying suspicious accounts that act as bots, whose results demonstrate the large role played by bots in key moments of recent Brazilian politics.

Although the bots operate in favor of specific agendas, that does not mean that they completely dominate the net or that the final perception of the majority of people will be a direct result of the influence of these devices. What we have found, however, is that they exist, they already operate on the Brazilian debate, they follow patterns and they seek to influence.  Above all, this research effort presented here wants to issue an alert that we are not immune and that we must be concerned with seeking to understand, filter and report the use and dissemination of false or manipulative information through this type of strategy and technology. We must be attentive and protect the democratic spaces, including on the social networks.

Considering that the upcoming elections will have a critical importance for the country, and supposing that our case will not be so different from other democracies in recent electoral periods, where clear manipulation attempts occurred (France, United States etc), we demonstrate with this effort two of DAPP’s commitments. The first one is related to the monitoring of the debate on the networks and to the attention to democracy. The second one is the continuous effort to develop and improve technologies to detect and understand this phenomenon.

The first phase of this study, presented here, was concentrated on political moments of high repercussion on the networks in the past three years: the elections of 2014, the impeachment of Dilma Rousseff, the municipal elections of 2016 and the general strike of 2017. The analysis considered multiple characteristics and metadata that indicate the presence of suspicious accounts.

The study of the use of bots in the period analyzed already demonstrates clearly the damaging potential of this practice for the political dispute and the public debate. One of the most evident conclusions related to this issue is the concentration of these acts in political poles located in the extreme end of the political spectrum, artificially promoting a radicalization of the debate and, consequently, undermining potential bridges for dialogue between the different existing political fields. Another glaring element is the “swelling” of political movements that are, in reality, of a much smaller dimension. When added together, these and other risks represented by bots are more than enough to cast light on a real threat to the quality of the public debate in Brazil and, consequently, of the political and social process that will define the years to come.


Marco Aurélio Ruediger
Diretor da FGV/DAPP