In This Issue
Frontiers of Engineering
December 1, 2008 Volume 38 Issue 4
Winter 2008 issue of The Bridge on Frontiers of Engineering

Driving Attention: Cognitive Engineering in Designing Attractions and Distractions

Monday, December 1, 2008

Author: John D. Lee

A driver's attention is a limited, critical resource that can be compromised by distractions.

Driving confronts people with many of the same demands as other high-tempo, high-consequence, complex activities. People who provide health care, manage power plants, and control aircraft face similar multitasking demands, many of which are mediated by technology (Hollnagel et al., 2006; Moray, 1993; Vicente, 1999). Drivers must divide their attention among navigation, hazard detection, speed control, and lane maintenance. In addition, drivers often engage in non-driving activities, such as conversing with passengers and adjusting entertainment systems. In this multitask situation, a driver’s attention is a limited, critical resource, and safety can be compromised when a driver fails to direct attention to the right place at the right time.

A recent study based on detailed data on 100 vehicles for a year showed that distractions and inattention (e.g., fatigue) contributed to approximately 80 percent of crashes and that distraction contributed to approximately 65 percent of rear-end crashes (Klauer et al., 2006). Unfortunately, this problem is likely to get worse, because driver distractions are likely to increase with rapid advances in wireless, computer, and sensor technologies (Regan et al., 2008). Not only will drivers have to manage cell phones, radios, and CD players, but they may also be tempted to use text messaging, select from MP3 music catalogs, and retrieve information from the Internet. Rapid changes in vehicle design are being made to accommodate these new devices. Nearly 70 percent of new 2007 vehicles are compatible with MP3 players, and all 2009 Chrysler vehicles will have wireless connections to the Internet (Bensinger, 2008). These infotainment devices have the potential to make driving time more enjoyable and productive, but they also have the potential to distract drivers.

Sensor, data fusion, and control technologies promise to improve driving safety by mitigating the distraction potential of infotainment devices. Increasingly, vehicles are being equipped with sensors that monitor surrounding vehicles to identify potential collisions, warn drivers, and even respond with emergency braking. Similar technologies that can automate driving during routine situations include adaptive cruise control that accelerates and decelerates the vehicle to maintain a constant speed or constant distance from the vehicle ahead (Walker et al., 2001).

Other devices can assist drivers with emergency braking, help them keep the car centered in the lane, and attend to potential threats of collisions (Norman, 2007). Although these developments are promising, driver-support technologies may not deliver the promised safety benefits because (1) they often respond imperfectly and (2) they may encourage people to pay less attention to driving if they think the system will protect them from distraction-related lapses (Evans, 2004; Stanton et al., 1997).

Figure 1 Complementary capacities of technology and humans.

FIGURE 1 The complementary capacities of technology and humans. When properly integrated, the combination is more effective than either of them alone. When poorly integrated, the combination is less effective than either of them alone.


As new technology has done in other domains, the introduction of infotainment and driver-support technology will fundamentally change driving. The complex array of factors that affect driving safety means that focusing simply on improving technology (e.g., designing a more capable automatic braking system) will not ensure that driving is safer, not only because technology will remain imperfect, but also because safety ultimately depends on leveraging a driver’s capabilities. Technologies must be designed in a way that attracts a driver’s attention to what matters most and does not annoy or distract a driver from safety-critical events.

Figure 1 illustrates the challenges of combining people and technology. The top diagram shows the complementary capacities of humans and technology—both are limited and may overlap to some degree. The middle diagram shows an effective combination of human and technological capacity—in combination, both perform better than either does alone. The bottom diagram shows a dysfunctional situation in which combined human/technology performs worse than either does alone; this can occur if the person does not capitalize on the capacity of the technology (on the left) or relies on the technology inappropriately (on the right). The disuse and/or misuse of technology often occurs when a new technology is introduced (Parasuraman and Riley, 1997). In addition, some technologies, such as warning devices, can annoy people and undermine product acceptance (on the left). Poorly coordinated technology can also interfere with a driver’s ongoing response to a situation (on the right).

Achieving an effective human/technology combination requires a deep understanding of how technology mediates human attention and decision making (Lee, 2006). The dynamics of attention can be considered as a multilevel process (Michon, 1989; Sheridan, 1970). At the operational level, attention is modulated over a span of milliseconds to seconds; at the tactical level, modulation may take many seconds or minutes; and at the strategic level, it may take hours or even months. Technology can have a powerful influence at any of these levels.


Figure 2 Technology-mediated attention.

Figure 2 shows the dynamics of how technology influences attention to driving and competing tasks through feed-forward, feedback, and adaptive control. With feed-forward control, drivers and technology anticipate upcoming demands and direct attention accordingly. Feedback control directs attention according to the evolving demands of the situation. Adaptive control directs attention based on changing goals and priorities. As technologies become more sophisticated and ubiquitous, they also increasingly influence drivers at all levels of attention and for each type of control.

Augmentation Rather Than Automation
Cognitive engineering is engineering with a sensitivity to human cognitive characteristics to improve safety, performance, and satisfaction. For example, rather than using technology to automate an action in an effort to eliminate human error, a more bene-ficial approach, and one that may yield greater safety benefits, would be to augment, rather than automate, human capabilities.

Technology makes it possible for a vehicle to monitor both the roadway and the driver. Thus it could augment the driver’s awareness of the roadway conditions and improve the driver’s awareness of his or her capacity to respond to those demands. Technology might improve safety by measuring the degree to which the driver is distracted and then directing a distracted driver’s attention by alerting the driver to roadway demands. In the following descriptions of how emerging vehicle technologies might mediate a driver’s attention, the reader should keep in mind that similar approaches might also apply to other high-tempo, multitask activities.

Using Model-Based Distraction Estimates to Improve Self-Awareness
In a survey of 1,000 drivers, 80 percent said they thought they drove more safely than the average driver (Waylen et al., 2004). This sense of confidence and, perhaps, complacency is one factor that encourages drivers to divide their attention between the roadway and infotainment systems. Augmenting a driver’s awareness of his or her attention to the roadway might be an effective way of mitigating distraction and helping drivers make better decisions about if and when they can safely engage in a distracting activity.

Estimating the degree of distraction experienced by a driver may be critical in helping that driver manage distraction. Figure 3 shows the output of a model of a driver switching attention between the roadway and an in-vehicle device (Hoffman, 2008). This model is based on dynamic field theory (Erlhagen and Schoner, 2002) and captures the time-varying factors that cause drivers to persist in looking away from the roadway (e.g., task inertia) and factors that draw a driver’s attention back to the roadway (e.g., increasing uncertainty about the roadway situation).

Figure 3 Motor Planning
FIGURE 3 A theoretical approach to describing the dynamic distribution of attention between the roadway and an in-vehicle device.

The top-down, or model-driven, estimate (described above) of how drivers distribute their attention can complement a bottom-up, or data-driven, approach to estimating a driver’s state based on real-time driving performance data. Bayesian networks and support vector machines are effective data-driven techniques for estimating distraction based on eye movements and steering behavior (Liang et al., 2007, in press). Increasingly instrumented vehicles provide an enormous volume of data that can be used as feedback to drivers and designers, provided those data are interpreted correctly.

Estimates of impairment related to distractions, such as text messaging, can augment a driver’s awareness of impairment in three ways (Donmez et al., 2006, 2007). First, a model-based prediction of distraction could alert a driver to upcoming conflicts so that he or she can direct attention to the roadway proactively. Second, the history of distraction and the associated decrements in driving performance could be shared with drivers after a drive to help them calibrate their own estimate of how well they can manage distractions. A third approach takes into consideration the current state of the driver when redirecting his or her attention to demanding roadway situations. This approach is described in the following section.

Alerting and Informing a Driver to Enhance Roadway Awareness
Sensor and algorithm technologies have made it possible for a vehicle to detect hazards and alert or inform the driver, thus reducing his or her reaction time to an imminent collision (Lee et al., 2002). Unfortunately, these systems also generate many false alarms—signaling a hazard where none exists—which can annoy and distract drivers. However, making such systems more useful and trusted will require more than a techno-logical fix.

For example, based on our knowledge of human reactions, we know that drivers perceive seat vibrations as less annoying than auditory alerts (Lee et al., 2004). In addition, not all false alarms are created equal. False alarms that drivers associate with events in the environment lead them to trust the system and thus become more likely to comply with subsequent alerts. False alarms that appear as if they occur randomly tend to have the opposite effect (Lees and Lee, 2007). Drivers respond differently to alerts, even though they might all be labeled “false alarms” from a technological perspective.

Adapting a threshold for alerts based on the degree of driver distraction could reduce false alarms by raising the threshold for attentive drivers. This approach could lead to an interesting paradox in that the drivers who most need alerts are also the most likely to consider them false alerts. For example, a distracted driver might not notice a hazard (even with the alert) and so might not appreciate the value of the alert. Providing a driver with information on roadway demands and hazards after a drive, similar to the post-drive feedback for distraction, could help him or her understand the reason for the alerts. More generally, drivers are more likely to benefit from vehicle technology that augments driver attention by informing through continuous information rather than alerting through discrete warnings.

Recent studies suggest the potential benefits of post-drive feedback (McGehee et al., 2007; Tomer and Lotan, 2006). In one study, teenage drivers drove with a camera that captured abrupt braking and steering responses. The resulting video and a summary of events was shared with their parents weekly, leading to an 89 percent reduction in the number of events triggered by risky drivers compared to the baseline period. Even after the feedback was removed, the rate of events remained low until the end of the study six weeks later. Whether feedback would be accepted or effective in helping experienced drivers manage distracting technology remains to be seen.

Technology changes the nature of driving by introducing new vulnerabilities and capacities (Woods and Dekker, 2000). Infotainment systems introduce new distractions that can undermine safety. Driver-assistance technologies promise to mitigate these distractions and improve safety. But we will not reap the potential benefits of these devices with a technology-only approach. Drivers tend to reject or misuse imperfect technologies that automate driving rather than augmenting driver capabilities. Cognitive engineering methods can show the way to using technology to leverage human capabilities to improve the safety and performance of complex systems by enhancing self-awareness and the awareness of potentially distracting technology.

Increasingly pervasive and powerful driving technologies, as in other domains, can blur the boundaries between the human and the technological, posing practical, theoretical, and philosophical issues about safety and performance, which increasingly depend on a complex interaction of driver, in-vehicle technology, and the driving situation (Lees and Lee, 2008).

Cognitive engineers face the following challenges:

• Philosophical issues relate to technologies that generally help but can also interfere with human performance. Driver-assist emergency braking, for example, generally improves crash outcomes, but, in rare instances, can impede a driver’s responses.

• Practical concerns include how to draw meaning from large, complicated streams of sensor data in real time and from petabytes of accumulated data to provide feedback to operators and designers.

• Theoretical concerns relate to the dynamics of attention and how technologies can affect those dynamics and, generally, how the nature of cognition changes as technology shapes and is shaped by human activity.

Much of the research described here was part of the SAfety VEhicle(s) using the adaptive Interface Technology (SAVE-IT) Program, sponsored by the U.S. Department of Transportation–National Highway Traffic Safety Administration (NHTSA) (Project Manager: Michael Perel) and administered by the John A. Volpe National Transportation Systems Center (Project Manager: Mary D. Stearns).

Bensinger, K. 2008. Chrysler will offer wireless Internet access in 2009 models. Los Angeles Times. Available online at, 0,1676276.story.

Donmez, B., L.N. Boyle, and J.D. Lee. 2006. The impact of driver distraction mitigation strategies on driving performance. Human Factors 48(4): 785–804.

Donmez, B., L.N. Boyle, J.D. Lee, and G. Scott. 2007. Safety implications of providing real-time feedback to distracted drivers. Accident Analysis and Prevention, 39(3): 581–590.

Erlhagen, W., and G. Schoner. 2002. Dynamic field theory of movement preparation. Psychological Review 109(3): 545–572.

Evans, L. 2004. Traffic Safety. Bloomfield Hills, Mich.: Science Serving Society.

Hoffman, J.D. 2008. A dynamic field theory model of shared visual attention while driving. Iowa City, Iowa: University of Iowa.

Hollnagel, E., D.D. Woods, and N. Leveson. 2006. Resilience Engineering: Concepts and Precepts. Burlington, Vt.: Ashgate.

Klauer, S.G., T.A. Dingus, V.L. Neale, J.D. Sudweeks, and D.J. Ramsey. 2006. The Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis Using the 100-Car Naturalistic Driving Study Data. DOT HS 810 594. Washington, D.C.: National Highway Traffic Safety Administration.

Lee, J.D. 2006. Human factors and ergonomics in automation design. Pp. 1570–1596 in Handbook of Human Factors and Ergonomics, edited by G. Salvendy. Hoboken, N.J.: Wiley.

Lee, J.D., J.D. Hoffman, and E. Hayes. 2004. Collision warning design to mitigate driver distraction. Pp. 65–72 in Proceedings of CHI 2004 Conference on Human Factors in Computing Systems. New York: Association for Computing Machinery.

Lee, J.D., D.V. McGehee, T.L. Brown, and M.L. Reyes. 2002. Collision warning timing, driver distraction, and driver response to imminent rear-end collisions in a high-fidelity driving simulator. Human Factors 44(2): 314–334.

Lee, J.D., M.A. Regan, and K.L. Young. 2008. Distraction as a breakdown of multi-level control at different time horizons. In Driver Distraction: Theory, Effects, and Mitigation, edited by M.A. Regan, J.D. Lee & K.L. Young. Boca Raton, Fla.: CRC Press.

Lees, M.N., and J.D. Lee. 2007. The influence of distraction and driving context on driver response to imperfect collision warning systems. Ergonomics 50(8): 1264–1286.

Lees, M.N., and J.D. Lee. 2008. Enhancing Information Acquisition from the Driving Environment. In Human Factors of Visual and Cognitive Performance in Driving, edited by C. Castro. Boca Raton, Fla.: Taylor and Francis.

Liang, Y., M.L. Reyes, and J.D. Lee. 2007. Real-time detection of driver cognitive distraction using support vector machines. IEEE Intelligent Transportation Systems 8(2): 340–350.

Liang, Y., M.L. Reyes, and J.D. Lee. In press. Non-intrusive detection of driver cognitive distraction in real-time using Bayesian networks. Transportation Research Board. Washington, D.C.: National Academies Press.

McGehee, D.V., M. Raby, C. Carney, J.D. Lee, and M.L. Reyes. 2007. Extending parental mentoring using an event-triggered video intervention in rural teen drivers. Journal of Safety Research 38(2): 215–227.

Michon, J.A. 1989. Explanatory pitfalls and rule-based driver models. Accident Analysis and Prevention 21(4): 341–353.

Moray, N. 1993. Designing for Attention. Pp. 111–134 in Attention: Selection, Awareness, and Control, edited by A. Baddeley and L. Weiskrantz. Oxford, England: Clarendon Press.

Norman, D.A. 2007. Design of Future Things. New York: Basic Books.

Parasuraman, R., and V. Riley. 1997. Humans and automation: use, misuse, disuse, abuse. Human Factors 39(2): 230–253.

Regan, M.A., J.D. Lee, and K.L. Young. 2008. Driver Distraction: Theory, Effects and Mitigation. Boca Raton, Fla.: CRC Press.

Sheridan, T.B. 1970. Big brother as driver: new demands and problems for the man at the wheel. Human Factors 12(1): 95–101.

Stanton, N.A., M. Young, and B. McCaulder. 1997. Drive-by-wire: the case of driver workload and reclaiming control with adaptive cruise control. Safety Science 27(2): 149–159.

Tomer, T., and T. Lotan. 2006. In-vehicle data recorder for evaluation of driving behavior and safety. Transportation Research Record 1953: 112–119.

Vicente, K.J. 1999. Cognitive Work Analysis: Towards Safe, Productive, and Healthy Computer-based Work. Mahwah, N.J.: Lawrence Erlbaum Associates.

Walker, G.H., N.A. Stanton, and M.S. Young. 2001. Where is computing driving cars? International Journal of Human-Computer Interaction 13(2): 203–229.

Waylen, A.E., M.S. Horswill, J.L. Alexander, and F.P. McKenna. 2004 . Do expert drivers have a reduced illusion of superiority? Transportation Research Part F: Psychology and Behavior 7(4-5): 323–331.

Woods, D.D., and S.W.A. Dekker. 2000. Anticipating the effects of technological change: a new era of dynamics for human factors. Theoretical Issues in Ergonomics Science 1(3): 272–282.


About the Author:John D. Lee is professor in the Department of Mechanical and Industrial Engineering, University of Iowa.