Roskomnadzor warns about the risks of advisory services
Algorithms that recommend content can spread fake information and mislead users.
Federal State Unitary Enterprise “Main Radio Frequency Center” (GRC), subordinated to Roskomnadzor, warns on the risks of developing advisory services. They can lead to the political polarization of society, the promotion of fake news and hate speech, follows from an analytical note prepared by the organization.
“Legislative regulation of recommendation services is necessary to protect consumers from non-transparent content selection algorithms that can distort the real state of affairs and lead to frightening consequences, up to real harm to human health and psyche,” the document says.
The work of recommender services is based on artificial intelligence technology, “therefore, it is safe to say that AI is a key component of popular social networks,” the note says. AI “often prioritizes contentious, hate speech, and misinformation because it is this type of content that grabs more user attention and leads to increased platform engagement and engagement.” As a result, social media can provoke political polarization and promote disinformation, the authors of the note said.
As an example, the authors of the document cite data from the United States: 64% of all new subscribers to extremist groups on Facebook (owned by Meta, recognized as extremist and banned in Russia) came to them through recommendations. Also mentioned is the search results monitoring system that the Republicans used to analyze recommendations on the Web on the eve of the US presidential election in 2020. According to these data, the researchers noted a significant bias in Google search results in favor of the Democrats.
Examples from the work of Russian advisory services in the GRFC are not given in the note.
Another risk of distribution of recommender services by the GRFC is that they give the user exactly the information that is similar to what he watched earlier. This leads to the appearance of the “echo chamber” effect, when the user is deprived of recommendations with an alternative point of view.
The GRFC proposes to introduce an audit of algorithms, a message for the user explaining how the algorithms work, as well as the ability to give and withdraw consent to the processing of personal data.