Safety implications of providing real-time feedback to distracted drivers
Introduction
Driver distraction can interfere with the driving task, by sharing resources required for driving activities, such as visual, auditory, motor and cognitive resources (Ranney et al., 2000, Wierwille, 1993). A distracting activity generally encompasses more than one of these components. The introduction of in-vehicle information systems (IVIS), such as navigational displays, raises concerns because of the greater potential to distract the driver. The conflict between IVIS and driving-task demands is a source of driver distraction (Verwey, 2000) and needs to be mitigated.
Feedback can help an operator learn how a system or the environment is changing (Vakil and Hansman, 2002) and is especially important in complex environments and systems such as observed in aviation, nuclear power plants, and driving. The driving environment can change very rapidly and the driver may fail to track these changes, particularly if the driver's attention is captured by a non-driving related activity or if the driver is cognitively loaded (Haigney and Westerman, 2001, Horrey and Wickens, 2006, Lee et al., 2001). In such situations, feedback can help the driver adhere to environmental changes more appropriately. It is important to note that feedback should not increase the cognitive load of the driver because concurrent feedback can interfere with ongoing task performance (Arroyo et al., 2006, Corbett and Anderson, 2001).
Previous studies have evaluated the potential benefits of different driver distraction mitigation strategies to alert the drivers of roadway events (Donmez et al., in press, Donmez et al., 2006a). These studies presented a taxonomy of distraction mitigation strategies, of which two have been evaluated: an advising strategy that warned drivers of a safety-critical roadway event (i.e. a lead vehicle braking or an approaching curve), and a locking strategy that stopped drivers from engaging in an IVIS during these same events. The study showed that a locking strategy was beneficial in improving driving performance during engagements in visual distractions (Donmez et al., in press). However, a disadvantage of these aforementioned strategies is the inability to effectively warn drivers of prolonged engagement in the IVIS. Some distractions may degrade driving performance to safety critical levels even on straight roads with low levels of traffic. Moreover, high levels of distraction can be an indicator of a possible performance decrement. Providing feedback when the driver is highly distracted can help avoid future hazardous maneuvers.
Another disadvantage of mitigation strategies based only on the roadway state concerns non-useful alarms. Although there is a roadway event, such as a curve, the driver may actually be focused on the driving task and be able to respond quite appropriately. An alarm provided in this situation can degrade system acceptance and result in frustration, which itself is a distraction that can have a negative effect on traffic safety (Burns and Lansdown, 2000). This can be avoided by giving drivers feedback based on their attentional state rather than just the roadway state. Moreover, compared to warnings based on roadway events, which can affect immediate performance, feedback on drivers’ engagement in distractions can generate a long-term behavioral change.
Research investigating ways to direct a driver's attention to the roadway is mostly focused on providing drivers with alerts based on roadway events, such as forward collision (Lee et al., 2002) or lane departure warning systems (Suzuki and Jansson, 2003). Although the concept of adaptive automation has been previously investigated (Parasuraman and Hancock, 2001, Rouse, 1994), the application in the driving domain has been limited. Emerging technologies, such as non-intrusive eye tracking systems, make it possible to monitor the level of driver distraction and adjust feedback provided to the driver accordingly. Such systems may be able to monitor the vigilance levels of the driver (e.g. an alertometer) (Knipling, 1999). Other systems may be capable of filtering incoming phone calls when a real-time workload estimate exceeds some threshold value (Piechulla et al., 2003).
A system that monitors distraction would ideally generate a continuous indicator that identifies when the driver is too distracted to respond appropriately. This indicator could be used to warn drivers, or even to take over vehicle control. There are several variables that can be used to identify the distraction level while driving (e.g. heart rate variability, eye movements, and driving performance measures). However, the nature of the task will have an influence on the appropriate measure. For example, off-road glances would not be a good indicator of auditory or cognitive distractions, but are quite useful for visual distractions. Off-road glances have also been widely used in the evaluation of distractions for non-driving tasks (Hoffman et al., 2005, Sodhi et al., 2002), driver experience (Underwood et al., 2003), and driver fatigue (Ji et al., 2004).
The location of the feedback may also influence its effectiveness. Visual feedback embedded in IVIS can be more effective for mitigating IVIS distractions since that is where the drivers’ attention is centered. However, visual feedback presented elsewhere may mitigate distractions other than those associated with the IVIS. For example, if the driver is talking to a passenger or viewing a map, visual feedback that is embedded in an IVIS may be less effective. Thus, different distractions may require different feedback mechanisms. Alternatively, one feedback central to the vehicle can be implemented to avoid the potential information overload due to multiple feedbacks.
This study investigates whether real-time visual feedback regarding drivers’ off-road glances can alter drivers’ interactions with IVIS and enhance driving performance. It is hypothesized that the distraction associated with visual tasks will undermine driving performance and that real-time feedback based on eye movements will lead drivers to glance away from the road less frequently and for shorter periods. The more frequent sampling of the road can result in better timesharing between driving and the distracting task, and therefore enhance driving performance. Two locations are tested for feedback in this study: IVIS-centered, i.e., incorporated within the IVIS, and vehicle-centered, i.e., displayed on the dashboard. It is hypothesized that IVIS-centered feedback will be more effective than vehicle-centered feedback.
Section snippets
Participants
Equal numbers of participants were recruited in two age groups: young (18–21) and middle aged (35–55). There were 36 drivers in the study. However, due to the reliability of the eye tracker, seven participants’ data had to be omitted. The final analysis consisted of 29 participants: 16 young (M: 19.5; S.D.: 0.9, 7 male and 9 female) and 13 middle-aged drivers (M: 43.6; S.D.: 5.5, 7 male and 6 female). All participants possessed a valid U.S. driver's license, had at least one year of driving
Results
The mixed linear model (PROC MIXED statement in SAS 9.1) was used to analyze continuous dependent measures. This model was set up to account for unbalanced designs (e.g. different number of subjects) and repeated measures on subjects where the residuals are not independent. The mixed linear model also enables a variance-covariance matrix to be adequately incorporated for repeated measures. There are different types of variance covariance matrices (e.g. compound symmetry, unstructured, first
Discussion
Feedback based on drivers’ off-road eye glances, was evaluated in a driving simulator at two display locations (IVIS-centered and vehicle-centered). Distractions had detrimental effects on driving performance. Regardless of display location, the feedback altered driver's interaction with IVIS resulting in less frequent glances to the display and longer roadway glances. Surprisingly, there were no significant benefits in their braking or steering response, even though drivers looked at the
Acknowledgments
This research was conducted as part of the SAVE-IT program under contract by Delphi Corporation and sponsored by the U.S. DOT – National Highway Traffic Safety Administration – Office of Vehicle Safety Research. The authors acknowledge the technical support and comments provided by Mike Perel of NHTSA and Mary Stearns of the U.S. DOT—Volpe Center. Data collection for this project was also provided in part by Gregory Scott working under the Iowa Biosciences Advantage (IBA) program.
References (42)
Differences in accident characteristics among elderly drivers and between elderly and middle-aged drivers
Acc. Anal. Prev.
(1990)Hazard and risk perception among young novice drivers
J. Safety Res.
(1999)- et al.
Drivers’ attitudes towards imperfect distraction mitigation strategies
Trans. Res. F: Psychol. Behav.
(2006) Young drivers’ overestimation of their own skill–an experiment on the relation between training strategy and skill
Acc. Anal. Prev.
(1996)- et al.
Changes in collision rates among novice drivers during the first months of driving
Acc. Anal. Prev.
(2003) - et al.
Extended time-to-collision measures for road traffic safety assessment
Acc. Anal. Prev.
(2001) - et al.
Reducing drivers’ mental workload by means of an adaptive man-machine interface
Transp. Res. F
(2003) - et al.
An analysis of driver's steering behaviour during auditory or haptic warnings for the designing of lane departure warning system
JSAE Rev.
(2003) - et al.
Approaches to mitigating complexity-driven issues in commercial autoflight systems
Rel. Eng. Syst. Safety
(2002) - et al.
A simple procedure for the assessment of acceptance of advanced transport telematics
Transp. Res. C
(1997)
A comparison of headway and time to collision as safety indicators
Acc. Anal. Prev.
Teenage drivers: patterns of risk
J. Safety Res.
A bayesian extension of the minimum AIC procedure of autoregressive model fitting
Biometrika
Developmental sources of crash risk in young drivers
Injury Prev.
CarCoach: a polite and effective driving coach
Locus of feedback control in computer-based tutoring: impact on learning rate, achievement, and attitudes
Assessing differences in young drivers’ engagement in distractions
Mobile (cellular) phone use and driving: a critical review of research methodology
Ergonomics
Visual sampling of in-vehicle text messages: the effects of number of lines, page presentation and message control
Transp. Res. Rec.
Cited by (156)
Exploring the Behaviour Change Wheel and the Theoretical Domains Framework in interventions for mobile phone driver distraction: A scoping review
2024, Accident Analysis and PreventionCar-following crash risk analysis in a connected environment: A Bayesian non-stationary generalised extreme value model
2023, Analytic Methods in Accident ResearchImpact of feedback content on dangerous driving behaviours based on the triangle model of responsibility
2023, Transportation Research Part F: Traffic Psychology and BehaviourAnalysis of distracted driving crashes in New Jersey using mixed logit model
2022, Journal of Safety ResearchAn Extreme Value Theory approach to estimate crash risk during mandatory lane-changing in a connected environment
2022, Analytic Methods in Accident Research