From discourse analysis to groupware design

Online collaboration has become ubiquitous. Remotely-taught educational courses, collaboration with work divisions in remote locations, and coordinating military personnel distributed across an information-driven battlefield all require design work to construct computer-mediated activities that enable participants to effectively coordinate remotely. Analysis methods that help system designers understand the complex interaction of online participants in a joint activity are crucial for designing effective software tools. This thesis describes how the referential structure of participant discourse can be used to model the interaction and recommend new system designs.; The thesis details a methodology for systematically analyzing online interaction, centered around applying ethnographically-informed analysis techniques to the discourse generated by participants. The methods we propose use references in the discourse as a way to infer the flow of information between representations. Experimental evidence will be presented showing that the methods produce models that reflect the reality of an interaction and predict the effect that changes will have on an interaction. Evidence will also be presented demonstrating the ability of the methodology to be taught and to be applied to a wide variety of domains. By matching the types of information, and the characteristic patterns of information access within the system, to specific representation properties, it is possible to recommend new representations which will improve task performance and reduce user frustration and error rate.