Decision Statistic Mapping and Number of Information Dimensions on Decision Making with Graphical Displays

Abstract

This study sought to: (1) analytically separate the components of a graphical display which contributed to performance on integrated and separable tasks; and (2) determine the effect of the number of dimensions of information which had to be integrated. To that end, the study employed a 7 × 3 mixed design with seven displays manipulated between-subjects and the number of information dimensions (three, six, and nine) manipulated within-subjects. The seven displays examined included two bar graphs (non-object and object formats), two midline displays (non-object and object formats), a direct graphical display, and two numerical displays (numerical separable and numerical integrative). Based upon propositions generated from emergent feature theory, the ability to integrate information in these displays should be a function of the faithfulness, saliency, and directness of mapping the decision statistic onto the display. Results indicated that the displays which directly represented the integrated decision, the numerical integrative and the direct graphical displays, resulted in the best performance. Intermediate performance was obtained on those displays (i.e. the object bar graph, the non-object midline, and the object midline) which incorporated faithfulness, saliency, or both, respectively. The worst performance on the integrated task was exhibited for those displays (i.e. the numerical separable and the non-object bar) which did not represent directness, faithfulness, or saliency. For both the integrated and separable tasks, accuracy increased as the number of information dimensions increased. The unexpected direction of this effect was attributed to subjects” investing more resources in performing the task at the six or nine cue levels due to the perceived increase in difficulty of the task.