On Weighted Egocentric Graphs and Social Group Communication

Abstract: Electronic communication has a profound impact on our society, where for instance social media now is almost ubiquitously used to share information. This has led to several major challenges, including how to overcome information overload and privacy concerns, how to utilize social context based on multiple data sources from both the virtual and physical worlds, and subsequently how to improve group communication. This doctoral thesis presents an Aggregated Social Graph (ASG) framework that utilizes weighted egocentric graphs representing the calculated social strength between people. The framework is based on a unified interaction model that support capturing and aggregating information from communication services such as social networks, mobile phones and email clients. Two algorithms for social strength computation are presented and evaluated in this thesis: Utility Function and Euclidean Distance. The proposed framework contains a social recommender engine that includes context-based methods (tags, locations and objects) to rank, filter, recommend and group social contacts and information based on the weighted egocentric graphs. The ranking of contacts can for instance be used to automatically form contextual groups. These groups can be used to dynamically compose and tailor communication tools for specific communication contexts by integrating widgets into web-based collaborative environments.The framework also contains a novel social distribution mechanism for controlling otherwise flat or viral distribution of information. The mechanism combine weighted egocentric graphs and users' context to establish a level of trust to control propagation of information, thus reducing the potential perception of spamming. The work is evaluated through several proof-of-concept prototypes that show the potential to improve distribution and filtering of information as well as social group formation. The proof-of-concept prototypes also show that communication tools can be dynamically composed for a specific group of users and for a specific context. Moreover, this thesis presents evaluation studies that compare the social strength algorithms and verify the concept of contextual group formation. In conclusion, utilizing social context as represented by weighted egocentric graphs has the potential to improve group communication services. We believe that the proposed framework is an effective means to reduce the problems with information overload and to enable automation of processes related to electronic group communication.