Modeling User Perception of Interaction Opportunities in Collaborative Human-Computer Settings

Interruptions are important for effective collaborative work, because agents often possess information required by others on their team. This need to get information from another agent arises in mixed human-computer teams as well as in homogeneous computer-agent environments. For example, a (human) driver may see changes in weather conditions that affect route selection while an automated navigation system without sensors does not. The navigation system may need this information to identify the best route. It is crucial to time interruptions, which are inherently disruptive, appropriately. Efficient interruption timing improves task performance as well as emotional state and awareness of the user, and decreases the negative effects of interruption (Adamczyk & Bailey 2004). A key aspect of reasoning about interruptions in collaborative settings is the ability to accurately estimate the costs and benefits of the interruption so that the outcome of the interruption positively affects group task outcomes. Cost estimation has been investigated in prior work on interruption management (Horvitz & Apacible 2003), but this work presumes a benefit to the user of having information the computer system can provide. The benefits of interruption have been studied in the adjustable-autonomy literature, but that work focuses on when to turn control over to a person (Tambe et al. 2006). Few models have combined these two aspects into an integrated decision making mechanism (Fleming & Cohen 2001), and none have done so in the kinds of fast-paced domains we consider, i.e., domains in which agents are distributed, conditions may be rapidly changing, actions occur at a fast pace, and decisions must be made within tightly constrained time frames. Furthermore, almost no attention has been paid to the possible discrepancy between a computer agent’s calculation of the utility of the interruption and a person’s estimation of the usefulness of the interruption. The failure to estimate accurately may lead to a person rejecting the interruption, and thus to a missed opportunity to improve team performance, turning the interruption into an unnecessary disturbance. Our research proposes a new model for interruption management. This model aims to help maximize the efficiency