Driver Performance with In-Vehicle Information Systems

Peter Centgraf

October 1, 2003

 

Task Structure and Interruptibility:

Preliminary Literature Report

Introduction

The primary goal of this project is to provide recommendations to GM for the design of safe vehicle information systems. These recommendations are to be based on empirical knowledge of human attention and information processing limits as relevant to the driving task. In particular, we are interested in research into dual-task performance and interruption and its implications for the design of systems. To this end, we began a search of the existing psychological and HCI literature. This search was guided by a categorization of task switching behavior likely to occur during use of a vehicle information system, summarized in Table 1.

 

	Name	Current Task	Interruption
1.	Init	Driving	Initiate Information Task
2.	Check	Information Task	Check Road Conditions
3.	Resume	Driving	Resume Information Task
4.	Notify	Driving	Notification from Information System
5.	React	Information Task	Critical Road Event

Table 1: Task Switch Types

 

Each type of task switch has a distinct context, and different factors are relevant to understanding driver behavior in each of them. The first three types can be categorized as voluntary or user-initiated. In each case, the driver makes a choice to interrupt the current task in order to perform another. It is important to understand the factors that influence the timing and frequency of each interruption and the impact of these parameters on the driving task. The question is primarily one of user motivation, since the choice to interrupt or not interrupt is made on an ongoing basis by the driver as he or she uses the information system. Therefore, general research into task switching behavior and multitasking performance at a high level will be the most applicable. Design guidelines and empirical studies of user preferences for notification in desktop systems may provide some insight. However, it is unclear if the relationships between task structure and user performance in the desktop context will generalize to the driving context. There are likely to be systematic differences in behavior between the two environments because of the cognitive priority given to the driving task and the physical constraints of the vehicle. Research involving other time-critical tasks such as flight control may provide a better analog to driving behavior.

The Init and Resume types of task switch are very similar. Both types involve a switch from driving to some form of information task. However, different stages of the information task present different challenges for the user. When initiating a new task, the primary challenge is to navigate the interface to find and issue the appropriate command. Once the task is in progress, the challenge may be to recover the state before interruption, such as visual position and subtask goals, or to react to information provided in a display. Investigations into guided choice and command/control systems would be appropriate for the former, while the effects of structure in visual displays and in task goals would be more relevant to the latter. It should also be noted that the user has more control over the initiation of a task than the individual steps within it. This could affect user confidence that they have verified safe road conditions, which in turn could change user choices of when to self-interrupt. The task process proper may also be less familiar to the user than the initiation process, which affects the level of attention required for each stage.

The Check type of task switch is necessary to monitor road conditions and maintain lane position. Cognitive models of driving can help determine what specific behavior is necessary for safe vehicle operation. The key questions are what factors support safe monitoring behavior and what are the parameters on those factors. The same design issues which moderate behavior in the first and third types of task switch may also play a role in a driver's road monitoring behavior. The recommendations we produce must attempt to match the design of the information system to the required behavior.

The final two types of task switch, Notify and React, can be categorized as involuntary or environment-initiated. The Notify type is similar to the low-priority interruptions common in desktop computing environments. In this case a technological system generates information intended for immediate consumption by the driver. This may be a progress update, a response to an information request, or a system-initiated message. There are some obvious requirements for these notifications, such as preventing shifts of attention away from driving during critical situations. Design recommendations for desktop systems may be applicable if the recommendations take into account high-priority tasks. Recommendations of notification techniques for use in safety critical systems might also be appropriate.

The React type of task switch involves an event in the physical environment that requires the driver's immediate attention. Examples of such events would be flashing brake lights on a vehicle in front of the driver, an animal crossing the roadway, the appearance of an important sign, or a change in the condition of the road surface. Although this type of interruption will probably occur less often than the others, it is the most critical for safety. Because it involves an immediate reaction to outside stimulus, psycho-perception research on a low cognitive level is applicable. Under what conditions would a user notice important road events, and what conditions would cause them not to notice? What is the effect of audio/visual modality of an information task on a driver's ability to recognize potential road hazards? We also need to know how the information task affects the reaction time of the driver after they notice an event. Basic cognitive psychology research into dual-task environments would be most likely to answer this question. Information processing stage models can form a basis for reasoning about the performance implications of particular designs.

 

Method

Many of the articles described below were gleaned from the literature section of the website "interruptions.net", created by Ivan Burmistrov at Moscow State University. The site includes an extensive list of articles addressing all aspects of interruption relating to human-computer interaction. These articles were augmented with additional basic cognitive research by searching for specific topic areas in the PsycINFO and PsycArticles periodicals databases. Examples of search keywords used include "attentional blink", "psychological refractory period", and "visual onset". All articles were filtered based on their empirical content and their applicability to the driving task. Comments regarding the content and relevance of individual articles were recorded in HTML format and are summarized below. When available, a digital copy of the full article text is linked from the references page, which can be found here:

http://www-2.cs.cmu.edu/~bej/tsi/private/references.html.

Research Summary

... forthcoming shortly ...

Conclusions

... coming soon ...