|
 |
|
|
Knowledge sharing networks (e.g. Q&A sites)
|
|
|
While static web resources are adequate for answering certain types of questions, they fail in others becuse they cannot match the variety of problems individuals encounter, nor can they provide a social dimension. Instead of using search engines, millions of users flock to online question answering (Q&A) sites: some to ask questions, others to answer them, and still others just have a good chat. In this vein of research, my collaborators and I are interested in the following problems:
- Identifying experts using the post-reply structure of Q&A forums:
- Clustering forums by expertise network structure & exploring the benefits of individuals' focus
UofM press release
- Monetary incentives to participation
- Motivation in participation and the relationship between intermittency and quality
- Modifying Q&A interfaces to match users by expertise and preferences
While at HP Labs, I worked, with Eytan Adar, on the PeopleFinder2 project, to mine intranet and extranet documents to automatically extract the HP collaboration network and individuals' expertise based on the content of documents that contained their names. |
 |
|
|
|
Information diffusion in networks (e.g. viral marketing)
|
|
How much of a role do social networks play in the information that we are exposed to? In particular, how does Web 2.0 both enable social influence and allow us to measure it?
- The vast majority of content in the online virtual world Second Life is created by the users themselves. And roughly half of such content diffuses from user to user through the social network. We quantify this social influence, and use it to predict the growth in popularity of particular items. We also find that the early adopters are distinct from the influencers, and that influencers don't necessarily have that many more friends.
- Viral marketing has the potential to boost the long tail of online product offerings. We study the success of person-to-person email recommendations and find that niche products are more likely to experience a boost from viral marketing. Other factors at play: the number of recommendations previously exchanged between the same two individual, the number of recommendations received for the same item, the cost of the them, and other features of the recommendation network.
-
Blogs are sometimes described as echo chambers. In this pair of papers, we try to infer which blog is getting information from which other blog. Warning, blogs can be infectuous!
Implicit structure and the dynamics of blogspace,
Adar, E., Zhang L., Adamic Lada A., and Lukose R. M.
, Workshop on the Weblogging Ecosystem, 13th International World Wide Web Conference, (2004)
Tracking Information Epidemics in Blogspace,
Adar, Eytan, and Adamic Lada A. , Web Intelligence 2005, (2005)
- Can community structure affect information flow? We study this in the context of a corporate email network.
Information flow in social groups,
Wu, F., Huberman B. A., Adamic Lada A., and Tyler J. R.
, Physica A: Statistical Mechanics and its Applications, Volume 337, Number 1-2, p.327-335, (2004)
|
 |
|
|
Structural features of online social networks
|
|
|
|
|
Information flow in financial networks
|
|
|
|
