Social networks collect and store all kinds of information about us: the sites we visit, data on search queries, deindividualized data from correspondence. Many sites have built-in trackers Facebook, Twitter, VKontakte. These trackers monitor our actions in order, for example, to better target ads. But the results are applied not only for advertising. One of their functions is to detect similarities between users of social networks.
Social media is part of a larger ecosystem. For example, VKontakte and Odnoklassniki are part of the Mail.ru Group, while Instagram and WhatsApp are owned by Facebook. This means that the amount of data they collect is enormous. Such an array of information allows users to be grouped in the form of a graph, where the distance between them is determined by unifying factors that have different “weight”.
How it’s done
Graph construction is the main way to identify similarities between network users. A graph is an abstract mathematical object that is a set of vertices connected by edges. In our case, users are the vertices, and the edges are the links between them.
The specific structure of the Facebook algorithm was not disclosed, but the company has a search service Graph Search allows us to assume just such an approach.
This is what the social connections graph looks like
Based on the graph, the social network identifies groups of users with family, work, cultural and other connections. If they are not yet friends, Facebook will show these people in the “You may know them” section.
The work of the recommendation algorithm is not disclosed, but users complain that at times they are “shoved” of people chasing them on the net.
How to get into recommendations for everyone
By understanding how algorithms work, you can try to use them to your advantage.
- Tag other users in photos. Such activity is highly valued and helps to get into the top recommended.
- Encourage users to visit your page. Someone else’s activity gives more “points” than yours. To get attention, like and write comments. Both Facebook algorithms and users will notice this.
- Avoid complaints. They are thrown out of the top for a long time.
What Facebook considers when ranking your friends (most likely)
Since social networks cannot yet track real life, the current ranking on Facebook raises many questions among users. The old algorithm trackedhow often users interact with each other, how long and how.
Modern – relies heavily on machine learning. As with recommendations, many indicators are taken into account. But it is reasonable to assume that the joint presence in photographs has much baboutmore weightthan likes or comments.
Here’s what most likely counts:
- User views of your page;
- Likes and comments on your posts;
- Presence on mutual photos or videos of mutual friends;
- Data from third-party apps (Facebook officially denies this approach, but this looks very believable).
We know what you want to hear
The results obtained strongly influence the news feed as well. If social networks showed everyone you follow, reading the feed would take much longer, and it would be almost impossible to get to the really important posts of people of interest to you on it.
Friends ranking is used to make life easier for users. Sorting users in the friends list by the number of messages in the correspondence (like in VKontakte) – only the top level. This is not enough to form a feed, because it does not reflect the importance of people in the user’s real life.
In network correspondence, we pay less attention to those with whom we constantly communicate live.
Friends are often so active that algorithms have to make a careful selection. As the administrators of the social network told, Facebook hides about 80% of our friends’ posts. In this case, the advantage is given to those whom the social network considers closer to the user.
Facebook’s algorithms are closely guarded secrets. But empirically, you can still study their features. So, in 2015, the journalist Megan Neal managed to find outthat when making recommendations, Facebook values our activity on other people’s pages more than someone else’s activity on ours.
Ranking posts and suggesting friends are sometimes criticized. For example, anthropologists see them as the cause of the “bubble filters»When the user is surrounded exclusively by information that corresponds to his views and interests (for example, political), which closes the opportunity to learn new things or revise their views. No less criticism is caused by the very possibility of “surveillance” in the pursuit of better advertising sales. However, these mechanics can also be used for good. So the Argentine rescue services managed to prevent the suicide of a woman who posted hints on the desire to commit suicide on the social network.
It is important not to forget that what you see on social media depends on your actions. This is not all the information, but what is selected specifically for you. If you learn to influence rankings, you can change the content of your feed. By methodically like and commenting on what you are really interested in, and not putting politeness likes under the posts of boring people, you can change your feed beyond recognition in just a few days.