Diffusion and Ranking in Digital Social Media
Social influence has always played a large role in the diffusion of information, the adoption of innovations,
and the propagation of recommendations. The rapid growth of digital social media such as blogs, Twitter,
and Facebook, has greatly expanded the reach and speed of such propagation and diffusion. Primarily,
it has done so by augmenting the ability of one individual to make their thoughts and opinions known to
a more extensive community, often beyond their immediate network and community. This same explosion
of online activity, enabled by these social web applications, and new computational techniques for the
analysis of social network data, has led to the growth of personalized recommendation and ranking systems.
These systems are capable of providing recommendations and identifying authoritative sources on any
number of topics from the next movie you should watch, to the next blog that will interest you. While it
may not initially appear that understanding diffusion and the development of authority flow based ranking
systems are connected, an exploration of both at the micro level - at the level of the individual - reveals that
both are fundamentally driven by the twin concepts of influence and authority. Understanding diffusion of
information is about understanding how individuals influence each other, while providing recommendations
is about understanding how to give influential advice on the basis of authority. In both cases, identifying an
authoritative individual as the source of influence or recommendation is critical.
In our research, we will collect and analyze multiple datasets from digital social media, tracking the
micro-level diffusion of information and flow of authority. Information diffusion occurs in different ways in
different formats, e.g., an unstructured web forum, the structured follower lists on Twitter, or a stream of
posts on a blog channel. Thus, it is important that the datasets that are collected are representative of a
variety of digital social media. Moreover, information about different topics diffuses in different ways. For
instance, reports on technology diffuse differently than news about popular culture, and opinion, in general,
diffuses differently from physical products. We expect that diffusion dynamics may differ both by platform,
as well as by topic area, within the platform. We plan to construct models that will allow for the examination
of different topics and platforms, taking into account both the sentiment of the message and the authority
of the transmitting individuals. To carry out this research, we will use a variety of approaches, including
agent-based simulation, network-based models, and authority flow based ranking.
This is joint work between William Rand, Louiqa Raschid, and Yogesh Joshi.