Effect of Reliability on Cue Effectiveness and Display Signaling

Twenty Army personnel using either a hand-held display (HHD) or a helmet-mounted display (HMD) were asked to detect, identify, and give heading information for targets hidden in a simulated mountainous terrain (the far domain) while performing a monitoring task in the near domain. Both displays had target cueing present for half of the trials, with the precision of the target cues varied across blocks. While all of the experimental blocks contained some trials with attentional cueing that was extremely accurate, placing the transparent cueing symbology on top of the target, some of the blocks had trials that also contained cues with degraded precision cueing, the cueing symbology located up to 22.5 degrees from the target center, or poor precision cueing, the cueing symbology located up to 45 degrees in visual angle from the target center. This manipulation simulated the effect of degraded sensor resolution. Explicit display of the precision reliability of the cues was attempted in order to help subjects diffuse attention during trials that lacked extremely accurate precision in cueing. During the last experimental block the automated target cueing catastrophically failed, causing users to experience costs with overtrust in automation, and then effects associated with the loss and potential recovery of trust in the automation for subsequent trials. Analysis of the results were conducted looking at two specific costs of attentional cueing that we define as type A (attention) costs and type T (trust) costs. Results show that spatially accurate cued targets were found faster than uncued targets. Targets were always found fastest on average under the cued condition while using an HMD, however, when targets were uncued, the salience of the different target types caused differences in the detection rate between the display platforms suggesting a scan/clutter tradeoff between HMDs and HHDs. Attention cueing induced a type A (attention) cost, shown by the low detection rate of a higher priority but uncued target when it was simultaneously presented with a lower priority cued target. For low salience targets, cueing that lacked spatial precision by more than 22.5 degrees of visual angle from the target center did not promote faster detection than if the target was uncued, however, the imprecise cueing remained beneficial for detection of the high salience targets even at the poor precision level. Type A and T costs were observed when the automation failed unexpectedly, with subjects initially showing signs of overtrust of the cueing information, and then on subsequent trials tending to undertrust the cueing information, with restored trust seemingly returning after a few reliable trials. Lastly, failures in automation seemed to mediate the effects of type A costs, as the detection rate of the higher priority but uncued targets increased.