Using derived social clusters in email to support high-throughput users in message management
Email is a story of both phenomenal success and frustrating challenges. The success of email as a professional communication tool clearly indicates that email is powerful and adaptable, but the success has led to the need for strategies to cope with the quantity and importance of email. The goal of this research was to develop a theory of how email is used, and from that goal, to develop a tool that will improve email interactions. The tool allows users to triage new mail, search existing mail, and categorize their email according to social clusters. These clusters are groups of senders/recipients with some association, as determined through collocation on multiple email messages.
The tool, SCuF (Social Cluster Filtering), is an extension of Microsoft Outlook 2003. It uses an extension of an existing algorithm (k-nearest-neighbor clustering) to derive a user-defined number of clusters of associated emailers. The user can then edit these clusters and use them to filter their inbox, another folder, or multiple folders. The filtered view only presents the user with messages to or from the members of the social cluster. The concept of social clustering extends the idea of social networks, and I present an analysis of extant work on social network analysis of email, showing that this dissertation research was novel and addressed problems that are not addressed by prior research.
This work was rooted in a theoretical foundation that draws from cognitive psychology and the semiotics of information. Cognitive psychology, and particularly cognitive and process models, allowed me to explore the cognitive mechanisms that people use when they interact with and make sense of the information contained within email messages. Semiotics of information allowed me to explore the visual and conceptual mechanisms that people use in these processes. Using analysis techniques from both psychology and semiotics, I created models of existing behavior and mechanisms, and extend these models to create a theoretically valid extension.
The models originated from fieldwork on how high-throughput email users actually use email. I focused on high-throughput users because they present a special case of intensive interaction with email. I provide data to define these users, and then present data from several in-depth interviews and observations of these users. Finally, I used high-throughput users in an evaluation of the tool.
0800: Artificial intelligence
0984: Computer science