Social Media Advertisements and the Problem of Disinformation

Abstract

Social networks have come to enhance our communication capacity and are important vehicles for the dissemination of ideas and news. However, fighting disinformation is like fighting an opponent. Based on what we learned from the 2016 American elections, we investigated a defense mechanism that could be implemented in the 2018 Brazilian elections. The idea was to create a system that would bring some transparency and allow us to audit the use of political advertisements on Facebook throughout the elections Brazilian companies. For this, we trained an algorithm based on deep learning, which proved to be able to distinguish political from non-political advertisements very precisely. When applying such an algorithm in our bank's advertisements, we found several election advertisements (political advertisements made during the election period) that were not in the Facebook database, that is, they were not declared as a policy by the advertiser. It should be noted that changes in terms of public policies to mitigate the problem of disinformation are happening little by little, but it is not possible to debate only the problem of disinformation on WhatsApp and leave loopholes on other platforms, such as Facebook.