Understanding and Augmenting Expertise Networks.

Abstract

This thesis investigates large scale knowledge searching and sharing processes in online communities and organizations. It focuses on understanding the relationship between social networks and expertise sharing activities. The work explores design opportunities of these social networks to bootstrap knowledge sharing, by using the specific social characteristics of social networks which can lead to sizeable differences in the way expertise is searched and shared. The potential impact of this approach was examined in three related studies using data from Java Forum, Yahoo Answers, and Enron.

The Java Forum study investigated how people asked and answered questions in this online community using advanced social network analysis metrics. Furthermore, it explored algorithms that made use of the network structure to evaluate expertise levels. It also used simulations to explore possible social structures and dynamics that would affect the interaction patterns and network structure in online communities. The Yahoo Answers study extended the Java Forum study into a more general community setting and covered much more diverse knowledge sharing dynamics. It analyzed both content properties and social network interactions across sub-forums with different types of knowledge, as well as examined the range and depth of knowledge that users share across these sub-forums. The Enron study, on the other hand, investigated how social network structure could affect the expertise searching process in organizational communication networks using simulations and social network analysis. Based on findings in these studies, a novel expertise sharing system, QuME, was proposed and developed.

This thesis provides a network theoretical foundation for the analysis and design of knowledge sharing communities. It explores new opportunities and challenges that arise in online social interaction environments, which are becoming increasingly ubiquitous and important. This work also has direct implications for practitioners. The ability to add the level of expertise would be a major step forward for expertise finding systems, and would likely open up a range of new application possibilities.