Evaluating the Visual Demand of In-Vehicle Information Systems: The Development of a New Method

Introduction

The last two decades have seen computer and communication technology become more prevalent in cars, enabling the incorporation of systems and devices for both driver support (e.g., navigation aids) and infotainment (e.g., news and email). These systems are generally designed to improve the safety, efficiency and comfort of the driver. There is growing concern amongst HCI specialists and Human Factors researchers that new forms of technology, accessible to drivers, could become significant sources of distraction and may have an impact on road accidents (Green, 1999a; Burnett et al., 2004; Pettitt, 2007). Although, distraction can occur as a result of events from either within or outside the vehicle, there is a real concern that poorly designed IVIS may introduce several dangerously distracting tasks into the driving environment (Pettitt, 2007).

In statistics from the UK, loss of control due to inattention was the most frequently reported contributory factor, involved in 35 percent of fatal accidents in 2006 (DfT, 2007). Although all forms of distraction could potentially negatively affect the driving task, visual distraction appeared to be a key contributing factor (Neal et al., 2005; Angell et al., 2006). It was reasonable therefore to infer that eye glance data contained important information for assessing the distraction effects of IVIS; in particular, glance duration. This has been noted by many authors to be an important ocular-based indicator of attention (Rockwell, 1988). Experimental research investigations into the degradation in primary driving tasks, as a result of sustained visual attention needed to achieve a secondary task, have generally attempted to highlight certain constraints, which secondary tasks must meet in order to be acceptable while driving (Gelau & Krems, 2004; Bhise, Forbes, & Farber, 1986; Green, 1999a). It is generally agreed that glances lasting longer than approximately 2.0 seconds are not acceptable while driving (Gelau & Krems, 2004).

Although a range of HF/HCI evaluation methods are available which allow researchers to describe, understand and predict relationships between variables in this area, practical concerns exist which limit their use in the real world. These include:

• Deciding Which Method is Most Appropriate at a Particular Stage of Development: Although some of the existing methods have been formalised e.g. occlusion (ISO, 2007), the majority of these methods could benefit greatly from further development (e.g., lane change test (LCT)) (ISO/WD, 2005), peripheral detection test (PDT) (Van Winsum et al., 1999). This lack of development often implies that there are no set procedures for implementing or assuring the reliability, and in some cases, the validity of the method. In this regard, designers and practitioners are often unsure as to which method would yield the best assessment at a particular stage in design.

• Deciding Which Methods are Most Useful: A further concern surrounds the measures obtained from the evaluation methods and their implications in relation to distraction. For example, methods such as field trials and simulator trials provide the opportunity to obtain direct measures of distraction. However, this involves frame-by-fame video analysis of IVIS interactions, which is a very expensive approach. On the other hand, methods such as occlusion (ISO, 2007), 15 second rule (SAE, 2000), PDT and LCT provide a less costly approach but predominantly focus on providing measures for visual allocation (glance frequency, total glance time, resumability (R)), vehicle control (lane deviation, deviation from a selected norm) and object and event detection (latency to detect, missing detection, situation awareness). These measures may imply degrees of visual and/or cognitive distraction but do not provide a direct measure of distraction. Practitioners are often unaware of which aspects of distraction are being measured.