Facebook knows its members very well. And Facebook is trying to bring them together. Under the heading “People you may know”, each member regularly receives suggestions for friend requests – now also on Instagram.
Work colleagues appear in the list next to party acquaintances or old school friends. In between, however, there are always completely unknown people – the selection of people is surprising at first glance. And we don’t even want to be friends with many of the people who are suggested to us. But how does Facebook know that we know these people? And according to what pattern are they selected?
This is how Facebook gets the data
Facebook draws its data from various sources to create the individual suggestions. The results are then calculated by a complex algorithm, little of which has so far been made public. One goal of the “People You May Know” section is to increase ad revenue on Facebook through more connectivity, more interaction, and longer visit times.
The company, now called Meta, does not reveal exactly how the algorithm works. On Facebook, however, the following is considered safe: Mutual friends, the overall structure of the previous friend network, data about education and work, Group memberships and the contacts that you (possibly unconsciously) imported from your smartphone address book to Facebook.
An absurd example of friend suggestions is known from the USA. Facebook wanted to connect the patients of a psychiatrist, even though they didn’t know each other. Even with the psychiatrist as a possible link, according to one “fusion“-Report no contact through the social network. One explanation for this would be that several patients had saved the doctor’s number on their cell phone – and Facebook allowed access to the smartphone contacts.
How exactly does the data comparison work?
Facebook, for example, compares the voluntary profile information with each other. Were members at the same school? Do they work in the same company? Did you grow up in the same place? If this information is not directly available, the software can also create a profile for the user through a few friend requests and, for example, suggest other work colleagues.
Facebook members disclose a lot of data voluntarily because they gave the company permission to do so when they registered. This includes, for example, personal telephone numbers and e-mail addresses. Facebook also uses imported contacts from the smartphone address book.
Another factor is the number of mutual friends. If two people have several mutual acquaintances, then they should also connect, says Facebook. For invitations and events that the person attends, Facebook also connects to other attendees and suggests them as new friends.
Link Prediction finds friends with whom there is no direct contact
In order to find friends with whom there is no direct connection, the Facebook algorithm uses the mathematical method of link prediction. Every “Like” information, every status message and every search input is included in the calculation. Facebook even uses everything that a member enters in the search bar for this purpose. EVERY interaction within Facebook is relevant for link prediction – even if it was just typed in and never sent. Since the Javascript code transfers all entered characters immediately, Facebook ALWAYS knows immediately – even if I delete my entry in search or chat before I send it.
Through all of these measures, Facebook can specifically suggest acquaintances as friends who users have met outside of the network. Even people who don’t really know anything about it. Sometimes it’s enough if two people are in the same place to have the other as a suggestion in the timeline.