Mechanisms of Controlled Sharing for Social Networking Users

Social networking sites are attracting hundreds of millions of users to share
information online. One critical task for all of these users is to decided the right
audience with which to share. The decision about the audience can be at a coarse
level (e.g., deciding to share with everyone, friends of friends, or friends), or at
a ne level (e.g., deciding to share with only some of the friends). Performing
such controlled sharing tasks can be tedious and error-prone to most users. An
active social networking user can have hundreds of contacts. Therefore, it can
be dicult to pick the right subset of them to share with. Also, a user can create
a lot of content, and each piece of it can be shared to a dierent audience.
In this thesis, we perform an extensive study of the controlled sharing prob-
lem and propose and implement a series of novel tools that help social net-
working users better perform controlled sharing. We propose algorithms that
automatically generate a recommended audience for both static prole items
as well as real-time generated content. To help users better understand the
recommendations, we propose a relationship explanation tool that helps users
understand the relationship between a pair of friends. We perform extensive
evaluations to demonstrate the eciency and eectiveness of our tools. With
our tools, social networking users can control sharing more accurately with less
eort. Finally, we also study an existing controlled-sharing tool, namely the cir-
cle sharing tool for Google+. We perform extensive data analyses and examine
the impact of friend groups sharing behaviors on the development of the social
network.