Supporting meaningful social networks.
University of Southampton, School of Electronics and Computer Science,
Recent years have seen exponential growth of social network sites (SNSs) such as Friendster, MySpace and Facebook. SNSs flatten the real-world social network by making personal information and social structure visible to users outside the ego-centric networks. They provide a new basis of trust and credibility upon the Internet and Web infrastructure for users to communicate and share information. For the vast majority of social networks, it takes only a few clicks to befriend other members. People’s dynamic ever-changing real-world connections are translated to static links which, once formed, are permanent – thus entailing zero maintenance. The existence of static links as public exhibition of private connections causes the problem of friendship inflation, which refers to the online practice that users will usually acquire much more “friends” on SNSs than they can actually maintain in the real world. There is mounting evidence both in social science and statistical analysis to support the idea that there has been an inflated number of digital friendship connections on most SNSs. The theory of friendship inflation is also evidenced by our nearly 3-year observation on Facebook users in the University of Southampton. Friendship inflation can devalue the social graph and eventually lead to the decline of a social network site. From Sixdegrees.com to Facebook.com, there have been rise and fall of many social networks. We argue that friendship inflation is one of the main forces driving this move. Despite the gravity of the issue, there is surprisingly little academic research carried out to address the problems. The thesis proposes a novel algorithm, called ActiveLink, to identify meaningful online social connections. The innovation of the algorithm lies in the combination of preferential attachment and assortativity. The algorithm can identify long-range connections which may not be captured by simple reciprocity algorithms. We have tested the key ideas of the algorithms on the data set of 22,553 Facebook users in the network of University of Southampton. To better support the development of SNSs, we discuss an SNS model called RealSpace, a social network architecture based on active links. The system introduces three other algorithms: social connectivity, proximity index and community structure detection. Finally, we look at the problems relating to improving the network model and social network systems.
Actions (login required)