VK talked about the role of algorithms to combat information noise on the Internet

Algorithms rank publications by relevance and take into account the individual interests of users. As a result, a person sees an orderly feed, rather than a chaotic stream of content. This was announced on June 16 by Alexander Poslavsky, head of the AI VK recommendations department.

Information noise is information that does not correspond to the interests and tasks of the user at the moment. As the volume of content consumed increases, it increases: there is more information than a person is able to consume, so without a filter it is difficult to identify what is really needed.

Modern recommendation systems are a chain of machine learning models that operates in several stages: from selection to final ranking. First, out of billions of pieces of content, it identifies the materials that are most likely to be of interest to the user. To do this, with a person's consent, algorithms analyze their views and interactions, as well as the content itself: format, subject matter, and content.

The system continues to learn from fresh data. If you rarely use the service, the algorithm relies on an outdated interest profile and cannot detect a change in preferences.

The algorithms then arrange the content in the appropriate order, depending on individual requests, the context of consumption (for example, time of day, frequency, and duration of sessions), and reactions to similar content.

Thanks to these steps, modern recommendation systems reduce the flow of irrelevant content. It is important that they do not lock a person in an "information bubble." The algorithm deliberately mixes in unfamiliar themes and formats. This is usually about 10% of the total recommendation traffic, which helps to combine the comfortably familiar with the inspiringly new.

On June 3, Stepan Kovalchuk, Senior Vice President for Media Strategy and Service Development at VK, spoke about the domestic market and the responsibility of large platforms to children. According to him, VK Video has a daily audience of about 8 million people around children's content, with significantly fewer users watching educational content than bloggers, TV series, and cartoons. Kovalchuk noted that it is important for the platform not only to keep the child's attention, but also to help him move towards more useful consumption.