Date of Award

Spring 2008

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Information Systems - (Ph.D.)

Department

Information Systems

First Advisor

Quentin Jones

Second Advisor

Starr Roxanne Hiltz

Third Advisor

Michael Bieber

Fourth Advisor

Brian Amento

Fifth Advisor

Brian S. Butler

Abstract

Millions of people use online synchronous chat networks on a daily basis for work, play and education. Despite their widespread use, little is known about their user dynamics. For example, one does not know how many users are typically co-present and actively engaged in public interaction in the individual chat rooms of any of the numerous public Internet Relay Chat (IRC) networks found on the Internet; or what are the factors that constrain the boundaries of user activity inside those chat rooms. Failure to collect and present such data means there is a lack of a good understanding of the range of user interaction dynamics that large-scale chat technologies support.

This dissertation addresses this gap in the research literature through a year-long field study of the user-dynamics of Austnet, a medium-sized IRC network (103 million messages sent to 7,180 publicly active chat-channels by 489,562 unique nicknames over a one-year period). Key results include: 1) the first rich quantitative description of a medium-sized chat network; 2) empirical evidence for user information-processing constraints to patterns of chat-channel engagement (maximum 40 posters and 600 public messages per chat-channel per 20-minute interval); 3) a short-term channel engagement model which highlights the extent to which immediate channel activity can be reliably predicted, and identifies the best predictor variables; 4) a model for the identification of factors that can be used to distinguish highly predictable channels from unpredictable channels; and 5) the first empirical study of how the Critical Mass theory can help in predicting the channels' long-term chances of survival by looking at their initial starting conditions.

Collectively, the results highlight how the knowledge of chat network dynamics can be used in making accurate predictions about the chat-channels' levels of short-term activity, and long-term survivability. This is important because it can lead to improved designs of future synchronous chat technologies. Such designs would benefit both the users of the systems, by providing them real-time recommendations about where to find successful group discourse, and the managers of the systems, by providing them vital information about the health of their communities.

Share

COinS