Attention and trust biases in the design of augmented reality displays.

摘要

This experiment seeks to examine the relationship between three advanced technology features (presentation of target cueing---and the reliability of that data, image reality, and interactivity) and the attention and trust provided to that information. In particular, we investigate the nature of two sorts of biases: an attention bias, in which the operator focuses attention to an area highlighted by the automation at the expense of other areas of the visual scene, and a trust bias, in which unwarranted attention is given to the guidance information.;Sixteen military personnel were asked to detect targets camouflaged in terrain, presented at two levels of scene detail (manipulated by varying the number of polygons with which the scene was generated and the level of the detail with which the scene was textured) while performing a terrain association task. Half the subjects actively navigated through the terrain; the other half passively viewed the control path of an active navigator. Cueing was presented for some of the targets, and the reliability of this information was manipulated at two levels (100% and 75%). More importantly, there were objects in the environment that were of high priority and uncued, and detection of these targets provided an assessment as to the degree of an operator's reliance on the cueing information. To assess trust in the terrain simulation, subjects were asked to perform terrain association given a hand-held paper map and report any inconsistencies between the visualization and the map.;The results revealed that the presence of cueing aided the target detection task for expected targets but drew attention away from the presence of unexpected targets in the environment. This attentional tunneling was mediated by cue reliability; unexpected targets, presented in conjunction with a cued object, were detected more often when cueing was only partially reliable. Neither image reality nor interactivity directly influenced trust in the display. Instead, the influences of interactivity were attributable to increased resource demand, which modulated performance in the presentation of unreliable cueing, and the effects of enhancing reality were attributable to the lower visibility of the target in the scene.