2015 \submitdate31 July 2015
The Emergence of Memes in Social Networks
Paul Kennedy and Thomas Osborn \examinerChairperson of Research Degrees Committee\firstreaderThomas Osborn
Abstract In recent years, research into social networks has rapidly progressed with the application of methods from the study of complex systems. Widespread user adoption of online social networks has created rich data sets regarding the structure and dynamics of social networks. The emergence of memes and viral content is frequently observed in online social media. However, research into complex, global and emergent phenomena is still an open area. The thesis of this dissertation is that the emergence of memes can be simulated in online social networks (OSN’s) by utilising multi-agent systems constructed to represent the network. This thesis is developed by first delivering novel methods for link prediction within online social networks. We show that link prediction can be used to classify the properties of a network, and also define the rule set for a multi-agent representation of the network. Using real world online social network data from Twitter and Facebook, we develop a prediction model for heterogeneous OSN users to engage with heterogeneous content. This model is then used to derive the rule set for a multi-agent system representing a real world OSN. We demonstrate how simulations of viral content and memes compare to real world cases. The results of this dissertation provide a better understanding of the complex dynamics influencing user activity within OSN’s. This research can better inform social marketing and advertising, deliver more effective recommendation systems in online social networks, and provide a basis for sociological theories with greater predictive power. \afterpreface
This thesis is focused on predicting the emergence of memes and viral phenomena within online social networks. Network effects play a strong role in the sudden and extensive propogation of viral content, so this aim is achieved by incorporating representative information regarding the network topology. To account for the diverse range of users of online social networks, as well as content and platforms, we adopt a multi-agent system approach.
There is a wealth of research already existing regarding social network analysis, which we review and consider in developing our method. The topology of social networks, in particular, represents a plethora of measures and techniques to classify and quantify structural properties. The diffusion dynamics across social networks has also received a great deal of attention, with many methodologies proposed. Finally, viral diffusion and the emergence of memes has attracted recent attention in the literature. However, there is still a research gap around the prediction of content and ideas diffusing in a critical, viral fashion, across a social network.