Recommending instead of taking down: YouTube hyperpartisan content promotion amid the Brazilian general elections

Abstract

YouTube released a global commitment to reduce the spread of problematic content by actively recommending “trusted” news sources on their platform but did not disclose the criteria used to classify the reputation of channels and the list of trusted news sources favored by the company. We aim to understand the patterns of recommendation by YouTube's algorithm and identify what the platform classifies as “reliable informative sources.” In our experiment, we performed 205 tests on YouTube's recommendation system, simulating the behavior of new Brazilian users with no historical data on the platform looking for reliable news throughout the elections. We documented the Brazilian media groups that were suggested on the homepage, further analyzing the presence, position and content of Jovem Pan videos, Brazil's flagship conservative media group. We identified that, far from reigning in extremist content, YouTube systematically prioritized Jovem Pan and did not actively take down toxic Jovem Pan content. Our findings indicate that YouTube magnified the imbalance between candidates' portrayals, exposing the regulatory asymmetry between broadcast media and online platforms in the country. In spite of the opaque commercial agreements and business strategies, the platform's recommendation decisions may have an “advertising effect” of sponsorship and endorsement on users.