Multimodal information presentation in support of timesharing and effective interruption management

Abstract

Operators in complex event-driven domains, such as aviation, face considerable attentional demands. They often need to time-share multiple tasks, and handle interruptions by other human and machine agents who may provide or request information in the interest of coordinating activities. The purpose of this study was to explore possible ways to support these timesharing and interruption management demands more effectively by distributing tasks across various sensory channels and presenting information on the nature of pending tasks to help subjects schedule their various activities. Participants performed a visually demanding air traffic control (ATC) task involving Data Link communication. At times, an interruption task had to be completed, which involved counting a subset of cues that were presented in visual, auditory, or tactile form. One group of subjects automatically received information about the nature of each interruption task whereas a second group was informed only about the presence of an interruption task. Within-subject variables included the modality/priority/frequency of interruption tasks and the workload level of ATC tasks. The results show that subjects delayed the initiation of visual interruption tasks significantly longer than the auditory and tactile tasks, which were less likely to interfere with the visual ATC task. When performed simultaneously with the ATC task, the visual interruption task led to the largest number of errors. Crossmodal interference was lowest when an auditory task was performed in parallel with the ATC task. Overall, findings from this study show that the presentation of concurrent tasks via different sensory channels leads to improved time-sharing performance, possibly due to reduced resource competition and scanning costs. Also, providing information about the nature of the pending interruption task helps participants schedule and manage interruptions more effectively. These findings will be discussed in terms of their implications for the design of multimodal interfaces in support of human-automation coordination.