A. Irune, G.E. Burnett

Pages: 21-30

Abstract
For in-vehicle information systems (IVIS) such as navigation and communication systems, designers must aim to understand user requirements early and evaluate these systems in an iterative fashion. Given that IVIS typically use complex visual displays, it will be critical to understand the visual demand of their interfaces and the subsequent effect on driver performance. System designers and car manufacturers require cost effective, valid, predictive methods to use in the design process of IVIS. Although methods such as the occlusion technique [1] lane change test [2] and peripheral detection [3] are available, they all possess drawbacks that hinder their practical use in the real world (e.g. need for proprietary software and hardware). In addition, none of the methods available provide a surrogate measure of glance duration. This is vital for identifying interfaces that encourage prolonged exposure to risk (i.e. long glances away from the road scene). Glance duration seems to be the most promising ocular-based indicator of attention. Rockwell [4] found there was a correlation between exposure to risk and average glance duration. Wierwille [5], Green [6], Tijerina [7], Farber [8], Greenberg [9] have suggested that, IVIS task should be accomplished in separate incremental steps with glance durations typically less than 2 seconds. A study was conceived to develop a preliminary approach to predicting secondary task demands based on primary task performance, with specific focus on glance duration. This approach draws on the visual alternating aspect of the occlusion method (i.e. switching from occluded to non-occluded periods) whilst incorporating a primary loading task (PLT) within the occluded period. By including a task in the occluded periods, the PLT approach aims to prevent people rehearsing their suspended goals, therefore ensuring the disruptive effect of the occluded cycle. Ten participants performed a range of secondary tasks (ST) while attending to a PLT. They were requested to alternate their vision between the PLT and the ST. The PLT, containing key components of the driving task (i.e. visual spatial search) occurred as part of a slide show presented on a screen monitor placed directly in front of the driver. The secondary task was performed on a separate screen monitor placed to the left of the driver, within the vehicle’s centre console. The secondary task was designed to promote glances with a wide range of durations. A positive relationship was found between the number of PLTs missed in a sequence and the mean ST glance duration (R= 0.01). These promising results highlight the potential for the use of PLT performance as a surrogate measure for glance duration. Further work is aiming to develop and validate this method by considering a range of real world in-vehicle tasks (e.g. satellite navigation task: destination entry, POI entry etc.) and comparing results with simulator trials.

Keywords: distraction; design; human-computer interface; human factor methods