Finding Social Circles in Ego Networks via Machine Learning
https://papers.nips.cc/paper/4532-learning-to-discover-social-circles-in-ego-networks.pdf
The article, “Learning to Discover Social Circles in Ego Networks” explores methods to use machine learning to group social circles together in one’s ego network. As discussed in class, an ego network is the network focused on one individual and all his/her friends, and the relationships between those friends. The algorithm that was used aimed to classify social circles by attempting to identify both similarities in members’ Facebook profiles and noticing the pattern of connections in the network. For example, a person may have a different group of friends from high school and college, with not many connections between the two groups. In this case, the algorithm would be able to pick up the social circle distinction by noticing the school that people went to from their profiles as well as the much lesser links and similarities between the two social circles. This analysis would bring rise to the presence of bridges, members that connect two social circles together in the ego network, which would clarify the distinction of social circles.
After testing the algorithm on over thousands of ego networks obtained from Facebook, the algorithm was found to rely on network structure far more than profile similarity in order to pinpoint social circles. This indicates that the network structure, itself, holds nearly all the information regarding an individual’s social circles, which was a concept reiterated in class as well. With the algorithm’s results, Facebook and other companies would be able to build on their privacy department: A users’ post would now be able to only be shown to those in his/her close social circles, and those who are relevant to the post. Furthermore, companies would be able to permanently hide a users’ posts from those individuals who share very minimal link to any of the users’ social circles, as those individuals are likely to be spammers. The ability to computationally unveil social circles from one’s ego network allows social media platforms to offer more cyber protection to users of the platforms.