Authoritarian News Personalisation and the 2021 Russian Parliamentary Elections

In this study we investigate how algorithmic news personalisation impacts dissemination of political news in the authoritarian context. To do so, we look at a major Russian news personalisation system - Yandex.Zen - which provides individualized news feeds to its massive audience (61.9 million users) and plays a key role in the regional media ecosystems, which the Russian government increasingly aims to control. Using as a case study the 2021 parliamentary elections in Russia, we conduct an agent-based audit of Zen-based personalisation system before, during, and after the election period to investigate what election-related news did it prioritize and whether this prioritization was subjected to bias (e.g., by prioritizing content favourable to the ruling party). Using four agent personas (e.g., pro-regime news reader or a news avoider), we identify the tendency of Zen to prioritize non-political content (e.g., hobby- and celebrity-related news) and content favourable to the ruling party on the eve of the elections, thus raising concerns about news personalisation serving as a form of hidden censorship in the Russian context.