Predictivity of system delays shortens human response time

System delays considerably affect users' experience and performance. Research on the psychological effects of system delays has focused on delay length and variability. We introduce delay predictivity as a new factor profoundly affecting user performance. A system delay is predictive when its duration is informative about the nature of consecutive interaction events. We report an experiment (N=122) where short delays were differently distributed across two alternative target stimuli in a choice response task. We manipulated variability and predictivity of delays. For one group of participants the delays were of constant duration. For three other groups the delays were variable, but differed in predictivity. They were either non-predictive, probabilistically predictive (they predicted the targets with a probability of 0.8), or deterministically predictive. Performance with constant delays was superior to performance with variable non-predictive or with probabilistically predictive delays. Surprisingly, participants with deterministically predictive delays outperformed participants in all other groups. This has important implications for interface design, whenever there is some degree of freedom in scheduling system delays. Best performance is achieved with predictive delays, but only when deterministic predictivity can be achieved. Otherwise, constant delays are to be preferred over variable ones.