Dissemination of Fake News in Cross-Platform Perspective: the Hydra of the 2018 Brazilian Presidential Election

Following international trends and other populist and polarized campaigns (2016 U.S elections and Brexit referendum), the 2018 Brazilian presidential election has also clearly illustrated the strength of false information circulating among online platforms, which ended up guiding the public agenda. This paper analyzes large-scale diffusion of fake news about the 2018 polarized Brazilian election from a cross-platform perspective. We adopt a digital methods approach with network analyses to understand the interconnections among WhatsApp, Facebook, Twitter accounts and public groups and how the viral information flows spread through out. The speed of these false narratives spread was central piece in a scenario of a populist and polarized campaign. But how was scaling up shaped in private networks without any public visibility, social information, exposure algorithm nor targeting applications in comparison with Facebook and Twitter? Methodologically, we have identified untruth stories with high spread level on each one of these platforms. In WhatsApp´s case, we monitored public groups dedicated to the political campaigns of six different candidates – giving special attention to the extreme right. The data collection included 500 thousand texts and images sent during five campaign months, allowing us to analyse the information flow and collective behavior of 90 interconnected political WhatsApp groups. In the Facebook and Twitter cases, we used a dataset that tracked 57 stories considered fake news by Brazilian main fact-checking projects. Adopting the number of sharings as a variable to measure their online relevance, we then selected and compared the most shared fake news on each platform. Some results on the fake news flow on WhatsApp suggest that the app is subjected to bipartite network logics due to its structure of interconnected groups and algorithm metrics can identify central groups in this time-varying process through stages. From stage to stage, fake news stories go from central nodes to peripheral ones exponentially amplifying its reach and going viral. Combining centrality (eigenvector) and identification of structured communities (modularity) provides an automated mechanism for detecting preferential starting points and routes more willing to make segmented fake news scale up in each community or coalition of communities found in a network.