Monitoring Automation Failures: Effects of Single and Multi-Adaptive Function Allocation

Abstract

Adaptive function allocation has been proposed to allow the advantages of task automation to be realized without some of the disadvantages of “static” automation. However, few empirical studies of the efficacy of adaptive allocation have been reported. The effects of adaptive function allocation on monitoring for automation failure during multi-task flight simulation were examined in two experiments. The first study examined the use of two methods of adaptive function allocation as a possible counter-measure to automation-induced monitoring problems. Subjects were required to perform a tracking and fuel management task while monitoring an automated system monitoring task for possible failures. For the “model-based” and “performance-based” adaptive groups, a single 10-minute block of fully manual performance on the monitoring task was allocated to subjects in the middle of a session. For the control group system monitoring was automated throughout all sessions. All three groups had low probabilities of detection of automation failures for the first 40 minutes spent with automation, before the adaptive function allocation change. However, detection probabilities were higher for both adaptive groups than for the control group following the function allocation change. The second study found that multiple adaptive changes (repeated function allocation) sustained these performance benefits over a longer automation period. These results clearly indicate that adaptive function allocation can improve detection of automation failures and that the improvement can be sustained over long periods of automation cycles.