How Does Facebook Determine Your Top Friends?
https://motherboard.vice.com/en_us/article/ezp4bj/facebooks-magic-formula-for-determining-your-9-top-friends
Over the last few years, the developers at Facebook have been adjusting and tweaking the algorithm they use to pair individuals and groups of people to determine their “closeness” as friends. Often times, users wonder how exactly Facebook “knows” who to put in their top friends list. Whether it be a close friend in real life with whom they never interact with online, or a complete stranger that constantly shows up in the “suggested” panel, Facebook’s algorithm leaves many questions as to how much information the company actually has on its clients at its disposal. To answer that question, Vice magazine’s Meghan Neal interviewed a representative from the company.
She reported that until recently, Facebook had been using an algorithm called Edge Rank which used social proximity, weight, and decay. Essentially, drawing weighted edges between nodes that grow stronger or weaker based on time. This builds on the social structures we have been dealing with in class, which are primarily characterized as “+” and “-“. However, the same principles would apply to this more complex network of nodes. Facebook’s current formula builds off of the Edge Rank system, including different kinds of weights for different actions in their platform such as tagging in a comment or a photo. Each of these actions strengthen the edge between two nodes, while inactivity weakens the edge. Another way to form weak edges, as suggested by the article, is by doing frequent searches of a specific user. While Facebook says “online stalking” via profile searches does not affect its algorithm’s edge rankings, tests done by Neal show a significant uptick in the frequency that those profile appear in her suggested friends after she had searched their names multiple times.
Facebook’s application of basic node structures between people into an algorithm that applies machine learning to grow and develop shows the applications to the theories we are learning in class. While the situations we are given in Problem Sets are far less complicated than those in the real world, the same principles of networks applies. Social media platforms such as Facebook rely on the theorems and principles (ex: Strong Triadic Closure) we are learning in this class to build their complicated algorithms to take data points and manipulate them to create a vast web of edges and nodes to connect their clients in the most efficient and practical manner. In addition, the fact that these algorithms are constantly changing and developing speaks to the importance of understanding the different applications of the theorems in this class.