Coterie: A Visualization of the Conversational Dynamics within IRC

Thesis -- August 10, 2001

Social patterns are observable in real-world interactions as visual cues. Online, however, there are few visual cues available that can be used to see and understand social patterns. In this thesis, I suggest that many of these social patterns are still present in our interactions online in text chat; they are merely encoded in the textual interactions. This thesis presents Coterie, a visualization of the conversational dynamics of an Internet Relay Chat (IRC) channel. Through Coterie, viewers can see the social patterns that underlie the text interactions between conversants. Using the chat messages posted to an IRC channel by users, Coterie builds statistical models for individual and channel-level interaction based on existing real-world sociometric models. Coterie also automatically separates out conversations using a conversation model based on a word usage algorithm. This information is then presented to the viewer through a novel display based on models for real-world small group interaction, which allows the viewer to see historical patterns of user interaction, such as a user's verbosity, as well as channel-level patterns, such as cohesiveness. The visualization is evaluated based on how well it makes such patterns visible, and further directions for its development are presented.

Available as: pdf

Movie available: quicktime 4

Software available for download (requires JRE 1.3) [125k zip]

Thesis Proposal -- December 11, 2000

Real-life social interaction depends heavily upon nonverbal social cues. We project many of these cues subconsciously, and they form the back-channels of our interaction with other people. Online, these nonverbal cues, especially those that are subconscious, are difficult or impossible to project to other people. This thesis proposes three related social interaction systems that augment a person's online representation with some of these social back-channel cues. These cues are generated and visually portrayed automatically from sociometric and historical interaction information. Similar back-channel generation and visualization subsystems and techniques are used within the three projects; each project applies these techniques to a different social situation. For example, a participant who is active within a social group may grow in size, or take on a brighter color, whereas a lurker may shrink and move off to the side of the group. The projects will then be evaluated based both on their visual design and ability to portray meaningful and useful social back channel cues.

Available as: pdf