Skip to Main Content
Our research addresses the question as to whether automatically collected quantitative data about people's behavior online can be analyzed to spot patterns that indicate behaviors of interest. Based on ethnographic studies, we find that people, going about their routine work, exhibit patterns in terms of their routine online activities and work rhythms. Such patterns can be comprised of many diverse types of events occurring over arbitrary durations. For example, they might include timing, duration and frequency of particular uses of hardware and software resources, manipulations of content, communication acts and so on. We use ethnography to identify and target significant patterns and computer logging to collect data on computer events that can be analyzed to find reliable correlates of those patterns. In this paper we discuss our methods and their potential for the development of novel types of applications that can identify normal activities and also spot telltale or deviant patterns. Such applications could be useful to users directly by providing helpful resources and content automatically or to the enterprise in general by automatically detecting performance problems, deleterious behaviors, or malicious activities.