A Framework and System for Intelligent Interruption Management

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

This research addresses one of the most important problems in human-computer interaction: interruption. E-mail notifications, agent requests, system alerts, and instant messages all periodically interrupt a user's tasks. Studies show that when tasks are interrupted at random moments, users take more time to complete tasks, commit more errors, and experience more frustration, annoyance, and anxiety than when tasks are interrupted at scheduled, opportune moments. In safety critical domains a small response delay or error committed due to poorly timed interruption can cost human lives or cause catastrophic damage, while in office settings increased frustration and anxiety hurt the user experience. When extrapolated over millions of users, the collective impact of interruption is quite remarkable. Nevertheless, users often desire or need the benefits that proactive systems provide, e.g., enabling near instant communication, maintaining awareness of information, being reminded of upcoming activities, learning to perform complex tasks, etc. The PI's basic hypothesis is that rather than interrupt immediately, proactive systems should wait for an opportune moment, that is to say a moment of lower mental workload such as at a task or subtask boundary, in a task sequence. In this research the PI will develop a theoretically-grounded framework and software that enable users to retain the benefits of proactive systems while mitigating negative effects of periodic interruption, thereby allowing users to be more productive and to experience less frustration, annoyance, and anxiety when multi-tasking. The framework will consist of empirically-based heuristics for assigning costs of interruption to various moments in a task, a description language for concisely expressing a broad range of user task behaviors, a task monitor that tracks execution of user tasks and builds a predictive model of execution, a decision algorithm that determines when to interrupt given a time window and a user's current position in a task, related tools, and empirical results and lessons. As part of the framework, the PI will develop attention manager software that defers granting requests for user attention until opportune (i.e., low cost) moments during task execution., yet which is useful even for requests with short timeframes due to urgency or relevance. The PI will empirically compare how well his attention manager balances awareness of information with mitigation of negative impact relative to other known strategies for interruption management. Broader Impacts: This research will refine and demonstrate a method for aligning mental workload to a model of task execution that will facilitate the use of workload as a new metric for evaluating complex interfaces. Project outcomes (including software) will be widely disseminated. The PI will furthermore integrate results from this research into a graduate course on human-computer interaction that offers multi-disciplinary perspectives on human attention; and will encourage participation from under-represented groups by funding at least one graduate assistant from an under-represented group for this research and by hosting two students from UIUC's Summer Research Opportunities Program each summer.