Testing a Computational Model of Interruptions: The Effects of Time Pressure on Interruption and Response Decisions

ObjectiveThe objective of this study is to empirically test a computational model of interruptions processes and effects, and to compare an alternative model to determine which best explains interruption and response decision making.BackgroundInterruptions in safety-critical environments (e.g., hospitals) can lead to an increased risk of error for the person being interrupted (the interruptee) but may be necessary for the person doing the interrupting (the interrupter) to maintain safety. Little research has considered the perspective of both the interrupter and interruptee.MethodWe tested a computational model of interruption and response decision processes through an experiment where participants (n = 312) worked as a nurse in a simulated clinical team. We examined how task progress, time remaining, and time pressure influenced decisions and compared the model with an alternative that allowed the effects of time pressure to be non-monotonic.ResultsUsing Bayesian hierarchical modeling, we found that a non-monotonic model best explained interruption decisions. Participants were biased toward interrupting, with time pressure only influencing decisions when it was very high. Contrastingly, the monotonic model best explained response decisions. Participants were more likely to accept interruptions as the interrupter's time pressure increased in comparison to their own.ConclusionTime pressure has a non-monotonic influence on interruption decisions, but a monotonic influence on response decisions.ApplicationFindings can inform interventions to consider the interruptions process from the perspective of both the interrupter and interruptee. Interventions could focus on training workers to more accurately assess time pressure when making interruption decisions.