U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-96-143
Date: April 1997
Development of Human Factors Guidelines for Advanced Traveler Information Systems and Commercial Vehicle Operations: Definition and Prioritization of Research Studies
CHAPTER 5. CONCLUSIONS
The goal of this task was to investigate some of the human factors issues specific to the acceptance of ATIS and CVO systems. This was first accomplished analytically by reviewing salient models and data on consumer acceptance of new products. Then three experiments were performed to collect new data, specifically directed at ATIS and CVO devices.
While there are several treatments in the literature of user acceptance of new technology, the models for this are quite complex and very difficult to apply directly to ITS. Nevertheless, these models were useful in suggesting initial directions for the empirical research reported herein. These experimental results were quite encouraging, both in suggesting additional research and for providing some tentative solutions to problems of driver acceptance of new technology.
A framework for driver acceptance of ITS technology is depicted in figure 81. It is based upon a synthesis and extension of the research in this report. The model starts (top of figure 81) with driver demographic characteristics (e.g., age, gender) and driver mental constructs (e.g., trust, tolerance). Any framework for driver acceptance of technology must start with the driver, not with the technology. To be commercially successful, ITS must focus first upon driver needs, and not upon technological capabilities. Technology is used when it fulfills a consumer need or want. The driver's needs and wants are conceived from the driver's mental model of the entire driving system: the vehicle, the roads, the driving environment including other vehicles, and the purpose of the journey. Technology is accepted when its comprehended benefits exceed its perceived costs.
These driver physical and mental characteristics determine feature pattern desirability (experiments 1 and 1B): the psychological structure that defines and differentiates elements of ITS functionality. Specific ATIS features can be divided into preference categories. For example, some desirable TravTek features are congestion information and route guidance to correct a route after a missed turn (table 3). Some undesirable TravTek features are vehicle position provided by voice and advertising information provided by voice (table 4). These components of the framework have been demonstrated in the present experiments. Figure 81 also implies that specific features are more likely to be accepted by drivers if they cluster into a factor (feature pattern). However, due to time and resource limitations this hypothesis was not evaluated in experiments 1 and 1B.
Future research is needed to better relate patterns of features to driver acceptance. Figure 81 hypothesizes that a new variable, termed ATIS complexity, is monotonically related to the sum of all ATIS features. Even neutral or unwanted features increase ATIS complexity. ATIS complexity decreases driver acceptance in two ways. First, it increases ATIS costs. Such costs include not only the dollar amounts needed to purchase and operate ITS technology, but also the added costs of using a more complex system such as increasing driver workload and effort to learn to use the system.
Second, ATIS complexity decreases driver comprehension of ATIS technology and its benefits. Desirable features increase driver comprehension and unwanted features decrease comprehension. Effects of neutral features could either increase or decrease comprehension but are anticipated to be weaker than effects of desirable or unwanted features. Driver acceptance is determined by comparing the benefits (driver comprehension) with the deficits (ATIS costs).
This framework predicts that driver acceptance will not be maximized by maximizing the set of ATIS features. Even if unwanted features are avoided, either by omission or by driver allocation of function decisions, too many neutral features will still increase ATIS complexity, thereby decreasing driver acceptance. Of course, if unwanted features are included, driver acceptance will diminish even more. If figure 81 is correct, it implies that good human factors practice would have manufacturers first test ITS features empirically for driver acceptance before including them in a production system. Having a feature that works well is no guarantee that drivers will find that feature to be desirable. More new technology is not necessarily better new technology. As was explained in chapter 1, even people who use new technology (e.g., ATM banking, VCR) seldom use all or even most of the features of new technology.
USE OF MODELS
The typical questionnaire study often reports results from individual questions, as was done in the Phase I analyses of experiments 1 and 1B, without any attempt at building an integrating model of acceptance, as was done in figure 9. This typical approach presents two problems of interpretation. First, it is difficult to integrate a large number of separate analyses of variance, often one statistical analysis per survey question; experiment 1 contained 52 feature–desirability variables. A more theoretical framework is required to better understand the large amount of data collected from questionnaire research. It is hard to grasp the true impact of such a large data set without such a framework. Second, the individual analyses fail to capture joint information about sets of desirable features. Without the Phase 3 analyses, it would not have been possible to understand that only six factors (table 11) are required to capture the desired feature patterns latent in the TravTek data set. Thus, the present research demonstrates the benefits of a model–driven approach to driver acceptance of ITS and strongly suggests that future research continue to use models.
A corollary of this approach is that global evaluations of a particular ATIS device may not be helpful for building design guidelines. While such evaluations may help compare different ATIS devices, knowing that some particular system is well–liked overall does not immediately suggest design improvements. Improvements can be generated from analysis of desired feature patterns, driver demographics, and mental model. A model is vital for detecting systematic departures from optimal human factors design.
ATIS AND CVO GUIDELINES
Designing ATIS Equipment
One goal of this task was to develop recommendations for the design of ATIS equipment. Experiments 1 and 1B showed that ATIS equipment should provide only the necessary or desired features, as unwanted features or even neutral features, will increase ATIS complexity, and reduce driver acceptance. This highlights the importance of user testing all potential features before system production. Experiment 2 suggests that some drivers may accept some unreliability in an ATIS system. However it should be noted that harmful inaccurate information influenced drivers differently than harmless inaccurate information. Experiment 3 suggested that commercial drivers favor those ATIS components that improve the driver's safety on the road. An introductory CVO system was recommended that included only those functions that are most directly and obviously supportive of safety. Given such an initial system, it was further suggested that other capabilities could be added incrementally as drivers came to appreciate both what the base system could do for them and what it could not do. It could also be that ATIS systems must be different for local drivers and for long–haul drivers, even in the introductory systems. Local drivers focused on a set of three ATIS functions that would improve their safety; whereas, long–haul drivers identified five safety–related functions. As with non–commercial drivers, both local and long–haul commercial drivers saw only negatives in some of the potential ATIS capabilities, such as broadcast services and an on–line services directory. Such irrelevant components were perceived as distractions at best.
An important goal of this task was to discover how consumers might be better educated to appreciate potential benefits (and costs) of ITS technology. Present results suggest that dynamic part–task simulators can help drivers understand ITS. The videotapes shown in experiment 1 were successful in helping drivers better understand the benefits of TravTek. Videotapes can be considered a low–cost part–task simulation. While drivers cannot exercise control in such simulations, they do learn from observation. Furthermore, direct experience with an ATIS feature, such as auditory instructions, alters driver acceptance of that feature. The CityGuide simulation of experiment 1B was also a successful educational technique. Drivers learned from observation but also could have been given an opportunity to interact more directly with CityGuide. While the CityGuide demonstration requires more equipment, it can be considered a low–cost simulator, also. The RGS used in experiment 2 allowed the greatest interaction because drivers were tested individually and could make route selection decisions on their own. Again, drivers were able to understand and appreciate potential benefits of ITS technology. These kinds of interactive "hands–on" simulators give drivers an opportunity to experience ITS directly at a detailed, practical, and individual level that cannot be achieved by reading a manual or lectures about the global benefits of ITS to society. Given current inexpensive PC–based simulators, this is a viable option. The educational approach taken for professional drivers may go beyond that used for the general public. A professional driver spends more time behind the wheel and may well benefit from a more extensive training and educational program. A more extensive part–task simulation would provide an opportunity for the professional driver to gain experience in the use of an ATIS function and to learn its strengths and weaknesses. A more emersive full–task simulation environment could afford professional drivers the chance to understand their ATIS systems in a non–lethal setting.
Incentives to Promote ITS Acceptance and Use in CVO
A final goal was to develop recommendations for incentives that could be used to promote ITS acceptance and use in CVO. Most people accept or reject products on the basis of individual experience rather than acting on a desire to achieve broader social goals such as decreased pollution. If social goals were more important to most individuals than personal convenience, many more people would use bicycles and mass transit to get to work. Thus, rather than statements of public approval about ITS, user benefits such as increased safety, shorter commute time, and better fuel economy should be emphasized. Experiment 2 showed that people will use an ATIS if it is provided to them. In fact, less than 8 percent of the subjects did not use the route guidance system when available. Despite this, it must be cautioned that drivers may resent paying money for inaccurate information and therefore may not use it. For commercial drivers, the incentive to accept ATIS systems lies in the increase in the driver's personal safety. If the first CVO system is limited to safety components, it should be accepted by drivers much more easily than a more complete functional system that is overly complex and intimidating and probably includes functions that appear to threaten the drivers' independence. Trucking companies also have a direct incentive to start with a simple safety–oriented ATIS. In recent years, the costs associated with accidents have ballooned either in the form of insurance premiums or in the out–of–pocket expense of self–insurance. Increasing driver safety may also mean a reduction in the accident rate. Another incentive for the trucking companies lies in possible reduction in driver turnover rates. A company that proves its emphasis on driver safety by fielding a safety–related ATIS system may benefit from increased driver loyalty thereby reducing its need to recruit and train new, often less experienced, drivers.
Establishing constructs, such as trust and tolerance, that mediate feature pattern desirability is only a first step towards creating human factors design guidelines. Highway engineers and manufacturers need quite specific guidance about parameters of constructs. While it is reassuring to know that drivers will tolerate imperfect systems and that errors do not completely eradicate trust, such general statements are insufficient for design guidelines. Design engineers need to know more precisely about human limits for trust and tolerance. Is 77 percent reliability a lower limit or will drivers accept 60 percent? These kinds of questions are best answered by more simulations of the type performed in experiment 2. Objective measures of behavior are superior to questionnaire ratings for obtaining design parameter limits.
Of course, it is not practical to obtain objective measures for all, or even most, design parameters. Questionnaire methodology casts a wider net for the same research dollar. Thus, an efficient research strategy would use questionnaires to determine which design parameters are most vital for driver acceptance of ITS technology, followed by objective measures to provide better estimates of key parameters. For example, experiment 2 used an overall level of reliability of 77 percent. But each driver did not experience exactly 77 percent errors, as each subject was allowed to choose an unique route. Future research could use dynamic selection of link traffic levels, which were fixed in experiment 2, to provide specified levels of unreliability for each driver on each simulated trip.
Future research must also take into account effects of attention and fidelity. These variables, although not shown in figure 81, alter the utility of research findings. Fidelity refers to the psychological fidelity of a simulator, not to its physical similarity to real devices. Behavior is controlled more by psychological fidelity, and often high physical fidelity represents an unnecessary research expense (Kantowitz, 1988). Attention controls how effective a simulation might be. For example, in experiment 2 decreasing the driver's bonus when heavy traffic was encountered helped to ensure that attention remained focused throughout the trip.
Commercial drivers may be the best initial group of drivers with which to assess the viability of ATIS functions, both at the concept/survey level and at the hands–on level. Commercial drivers typically operate larger, less easily controlled, vehicles and, therefore, would appear to have a greater need for straightforwardly usable and useful systems.
In summary, the tools and models developed in this task offer great promise for achieving the goals of this project. While evaluating consumer acceptance of future technologies is an arduous challenge, human factors does have methods to tackle the job effectively. These will be applied relentlessly for the remainder of this project.