Facebook Friend Suggestions and Social Networks
Companies like Facebook utilize networks to compile a list of friend suggestions for their users that can be surprisingly accurate. Facebook gets a good starting foundation of a user’s social network by compiling data from their synced contacts and location, which can become even more accurate when user’s input information on their profile such as hometown, places of education, place of work, etc.. This information helps tighten the network and reveal who may know who in a user’s social network, along with forming the separate connected components within the network. However, there are some flaws in the algorithm as gathering a person’s phone/email contacts, location, and other factors can’t necessarily inform the site of a strong or weak tie between the nodes (users) and edges (the relationships). This explains why users get friend suggestions from people that they know, but who they have no desire to connect with on Facebook. Facebook also tries to suggest people who a user may not know currently, but who could likely become a friend. They account for factors like degrees of separation with a person, number of mutual friends between users, along with other factors to try to predict probable future friendships.
Facebook’s friend suggestion algorithm connects to topics we’ve learned in class about social networks, specifically the strong triadic closure property. Facebook uses the strong triadic closure property (classifying a strong tie as a mutual friendship on Facebook) to suggest new friends who a user may not know currently, but who they have other mutual friends with. The strong triadic property in this case would determine that if A and B along with B and C are mutually friends on Facebook, A and C have potential for a future friendship if they don’t already know each other. It can become challenging to narrow friend suggestions when users have hundreds of friends, which has brought invited conversations about using more powerful technology such as artificial intelligence, to help tighten these social networks. This was proposed by using AI to track lifestyle similarities among users. This can be very invasive, as it stated in the proposition that this would require constant tracking of a user’s device even when they aren’t on the app to measure things such as physical activity to form commonalities. This creates questions about whether a higher level of technology crosses boundaries and invades too much user privacy or if these anticipated friend suggestions that would account for more similarities among users make it justifiable.
https://www.ijsr.net/archive/v4i12/NOV152104.pdf