Influence of Infrastructure Design on Distracted Driving (2022)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 Chapter 2. Distraction as It Relates to Infrastructure Elements Driver distraction has commonly been applied to scenarios where a driver physically engages in some activity, such as texting, that can divert his or her attention from the driving task. Since inattention is difficult to measure, distraction is often defined as glances away from the driving task. This definition is more challenging for identifying IRDs since glances toward the distracting feature would likely be within the same range of locations where a driver is focusing on the driving task. To address this challenge, Chapter 2 summarizes common definitions for distraction and explores how these definitions relate to infrastructure. Additionally, common surrogates for distraction, which may be applicable in the evaluation of IRD, are also discussed. 2.1 General Definition of Distraction A number of researchers, safety agencies, transportation agencies, and public health practitioners have offered definitions of distraction. For example, the World Health Organization offers the following definition: Driver distraction is generally thought to be different from driver inattention, or poorly allocated attention. Distracted driving occurs when some kind of triggering event external to the driver results in the driver shifting attention away from the driving task (e.g., a ringing mobile phone). Thus, the diversion in attention occurs because the driver is performing an additional task or is temporarily focusing on an object, event, or person not related to primary driving task. Inattention while driving applies to any state or event that causes the driver to pay less attention to the task of driving—the inattention can be present without necessarily having been triggered by an event, for example, daydreaming. The diversion of attention that occurs in distracted driving is also distinct from those impacts on driving performance that are attributable to a medical condition, alcohol or drug use, and/or fatigue (although these factors may compound the effects of distraction). External distractions may arise when the driver looks at buildings, people, or situations outside the vehicle, as well as at billboards and other roadside advertising (WHO 2011). Driver distraction is the diversion of attention from activities necessary for safe driving (Lee 2014, Regan et al. 2009). Much of the focus in safety research is visual distraction (i.e., eyes-off- road). Dewar and Olson (2007) define visual distraction as the momentary or transient redirection of attention from the task of driving to a thought, object, activity, event, or person. Visual distraction is defined as an activity that requires a driver to divert his or her gaze from the road to acquire information from the distracting source (Campbell et al. 2016). Cognitive distraction is a mental distraction that diverts a driver’s attention away from the roadway (i.e., mind-off-road) and may or may not involve visual distraction. Manual distraction is defined by NHTSA (2021) as anything that causes a driver to take his or her hands off the wheel. Visual distraction has shown the clearest link to traffic safety. As noted by Dewar and Olson (2007) and many others, the primary characteristic of interest is eyes-off-road time. Dewar and

8 Olson (2007) define eyes-off-road time as the sum of all time associated with glances not directed toward the road plus the transition time from eyes-off-road events. Other researchers have more broadly used non-roadway-related or non-driving-related glances, which excludes all roadway- or driving-related glances. Roadway- or driving-related glances include attending to the forward roadway and glances to the rear view mirror, console, or side mirrors (Birrell and Fowkes 2014, Yuan et al. 2018, Angell et al. 2006, Hallmark et al. 2015). In its most recent visual-manual distraction guidelines, NHTSA identified thresholds for visual- manual distraction related to in-vehicle secondary tasks, such as infotainment display features (Campbell et al. 2016). The purpose of these guidelines is to provide manufacturers with guidance about the design of in-vehicle tasks with respect to safety. The guidelines primarily rely on driver eye glance behavior as the metric of interest. Eye glance behavior refers to where a driver is looking and for how long. The NHTSA guidelines have two related thresholds for defining visual distraction with respect to glance behavior. First, a task should not result in any single glance away from the roadway lasting 2 seconds or longer within a 6-second window. Second, the total glance duration, or total period it takes for a driver to complete a task, should result in a total off-road glance duration (total eyes-off-road time) of 12 seconds or less. These thresholds were adopted principally due to their strong link to crash risk and safety. Data from several large-scale naturalistic driving studies demonstrate an increase in crash risk associated with both long (>2 seconds) single glances and long total eyes-off-road glance interactions (>12 seconds). For instance, in a 100-car naturalistic study, there was a significantly increased risk of a crash (odds ratio = 2.3) when a driver’s eyes were off the forward roadway for 2 seconds or longer (Klauer et al. 2006). Sixty percent of observed crashes involved at least one long glance away from the forward roadway. Similarly, in a naturalistic study with commercial vehicle operators, Olson et al. (2009) found an increase in crash and near-crash risk associated with eyes-off-road glances of 2 seconds or longer. In an additional analysis, Klauer et al. (2010) showed an increased risk of crash and near-crash events (odds ratio = 1.6) as eyes-off-road time increased. More recently, an analysis of data from the SHRP2 NDS supports these visual distraction thresholds. Victor et al. (2015) examined the risk of crash and near-crash events associated with long (>2 seconds) eyes-off-road glances. Long glances were associated with a significant increase in the risk of crash (C) and near-crash (NC) events, and the one-to-three-second window preceding the crash event was the best predictor of involvement in a crash or near-crash (CNC) event (see Figure 1).

9 Source: Victor et al. 2015, SHRP2. Figure 1. Odds ratios for eyes-off-road glances in windows within 12 seconds surrounding a crash. Victor et al. (2015) also found that eyes-off-path glance time in general was higher for sampled windows with a CNC event than for baseline (no-crash) windows. As noted, the topic of distracted driving within the human factors community has coalesced on visual distraction, which is an activity that requires the driver to divert his or her gaze from the road to acquire information from the distracting source (Campbell et al. 2016). A glance duration of two seconds away from the road has been shown to have a correlation to crash risk by several researchers (Klauer et al. 2006, Olson et al. 2009, Victor et al. 2015). 2.2 Identifying Driver Distraction in Crash Forms Misokefalou et al. (2016) observed that in research environments such as driving simulators and naturalistic driving studies, distraction is often presumed to have occurred based on outcomes such as crashes, near-crashes, unnecessary speed changes, lane departures, or sudden loss of vehicle control. However, since all of these outcomes also have other causes, the presumption of distraction is not satisfactory for crash reporting purposes. Similarly, a definition of distraction based solely on glance duration is not viable for post-crash observation and reporting by the attending officer. As a result, the identification of distraction in the context of most crash data depends on coding methods. Although all states include distraction as a contributing factor in crash report forms, practices and working definitions appear to lack consistency. The fourth and fifth editions of the Model Minimum Uniform Crash Criteria (MMUCC) include separate definitions for external and OVDs (NHTSA 2012 and 2017).

10 The fourth edition of the Model Minimum Uniform Crash Criteria (MMUCC-4), published by NHTSA in 2012, defines external distractions as “driver distractions that occur outside the vehicle, such as a crash in the next lane or on the other side of the median, automated highway signs, interesting objects in the sky, fire off the roadway, etc.” Separately, MMUCC-4 defines an outside-the-vehicle distraction as an event where “the driver was distracted by something outside the vehicle such as birds or other animals or a roadside fire. This may include unspecified external distractions.” Both definitions reference the same data field: P16. Notwithstanding the redundant terms and definitions, it seems clear that the intent of the MMUCC-4 data model is to include not only IRDs but also those caused by people, animals, or objects outside the vehicle, such as an action taken by another road user or someone near the road, a previous crash, or a natural phenomenon such as sunrise or sunset, lightning, aurora, etc. The MMUCC-4 data model provides a single data field for distractions, with the following enumerations: P16. Driver Distracted By Definition: Distractions that may have influenced driver performance. The distractions can be inside the motor vehicle (internal) or outside the motor vehicle (external). Attributes: • Not distracted • Manually operating an electronic communication device (texting, typing, dialing) • Talking on a hands-free electronic device • Talking on a hand-held electronic device • Other activity, electronic device • Passenger • Other inside the vehicle (eating, personal hygiene, etc.) • Outside the vehicle (includes unspecified external distractions) • Unknown if distracted The fifth edition of the Model Minimum Uniform Crash Criteria (MMUCC-5), published by NHTSA in 2017, includes the same seemingly redundant definitions of external and OVDs as its predecessor while implementing a more sophisticated data model. Specifically, MMUCC-5 defines two database fields to distinguish the source of a distraction from the resulting driver action. These data fields are part of NHTSA’s “involved-person data model” and are identified as “Distracted by Action” and “Distracted by Source.” Thus, there are two database fields associated with each person (driver or nonmotorist) involved in the crash: one describing the action indicative of distraction and a second describing the source of the distraction. In principle, these fields are used by responding officers to identify “distractions that may have influenced driver/nonmotorist performance, involving both an action taken by the driver/nonmotorist and the source of the distraction.” The data fields and enumerations are summarized as follows: P18. Distracted By S1. Action 00 Not distracted 01 Talking/Listening

11 02 Manually operating (texting, dialing, playing game, etc.) 03 Other action (looking away from task, etc.) 99 Unknown S2. Source 01 Hands-free mobile phone 02 Hand-held mobile phone 03 Other electronic device 04 Vehicle integrated device 05 Passenger/Other nonmotorist 06 External (to vehicle/nonmotorist area) 07 Other distraction (animal, food, grooming) 97 Not applicable (not distracted) 99 Unknown The logic checks recommended in MMUCC-5 do not allow a distraction with an “External” source to be combined with a “Manually Operating” action. Thus, the data model assumes that an external distraction will not provoke a phone call or manipulation of in-vehicle devices such as the radio, global positioning system (GPS), or heating and cooling controls. As a result, if the source is “External,” only three actions are considered valid for reporting purposes: “Talking/Listening,” “Other action (looking away from task, etc.),” or “Unknown.” Verifying that an external distraction has occurred appears to be a crash reporting challenge. While law enforcement officers sometimes discover physical evidence indicative of an inside the vehicle distraction (such as a cell phone with a partially written text message), physical evidence of an external distraction is often scarce. Law enforcement officers can be skeptical when a driver claims to have been distracted by something ephemeral, such as a deer running across the road. In other cases, a driver could have been distracted by an external stimulus but be unaware that a distraction has occurred. There are also cases where no one can provide investigating officers with information about whether external distraction had occurred (for example, if a crash kills or seriously injures all involved persons and there are no eye witnesses). In recent years, NHTSA has emphasized the need for making a clear distinction between distraction-related crashes and those related to inattention or fatigue (NHTSA 2018). 2.3 Driver Distraction from the Perspective of Infrastructure Elements 2.3.1 Transportation Features Related to Distraction Various studies have identified working definitions of IRDs. Most have simply focused on distractions that occur outside the vehicle. For instance, in a study by Salmon et al. (2011), a taxonomy of distraction sources for bus drivers was developed. The authors considered IRDs as those that include any features of the roadway infrastructure that the driver might find distracting. Examples included roadside advertising, lane width, and road layout or roadway signage.

12 In Driver Distraction: Theory, Effects and Mitigation, Horberry and Edquist (2009) defined a taxonomy of visual information in or near the roadway environment that may cause distraction. The authors noted that other nonvisual sources of distraction exist (e.g., noisy pavement, car sirens, agricultural or industrial odors) but focused only on visual distractions. Roadway factors the authors indicated as being distractions are summarized in Table 1. Table 1. Taxonomy of visual information distractions. Roadway Category Examples Built roadway • Road geometry (e.g., lane width). • Road surface. • Traffic signs and markings. Situational entities • Moving and parked vehicles. • Pedestrians on or near the roadway. • Weather (e.g., fog, rain). • Ambient light level (e.g., darkness at night). Natural environment • Trees and other vegetation. • Seas, lakes, and rivers. • Hills. Built environment • Houses and other buildings. • Bus stops. • Billboards and other roadside advertising. Source: Horberry and Edquist 2009. Using this taxonomy, Misokefalou et al. (2016) studied driver distraction factors on an urban tollway in Athens, Greece. The authors focused on distraction related to the built roadway environment and aggregated roadway elements into three categories: advertising, road elements or structures, and other elements. Included in the advertising category were billboards and gas station signing. Changeable message signs, noise barriers, informational signing, tolls, and bridges and overpasses were included in the road elements or structures category. The other elements category included buildings and railway stations. The study found that toll plazas, passenger rail stations adjacent to the roadway, variable message signs, and overpasses were potentially more distracting than advertising (Figure 2).

13 Source: Copyright ©2007 K. Krallis, User:Sv1xv, Creative Commons Attribution-Share Alike 3.0 Unported license Figure 2. Commuter rail station adjacent to the Attica Tollway in Athens, Greece. Transportation professionals have been concerned about the distraction potential of roadside advertising (e.g., billboards) for decades. The majority of studies related to infrastructure distractions continue to focus on roadside advertising, with recent studies noting the high distraction potential of digital billboards (Misokefalou and Eliou 2012a, Misokefalou and Eliou 2012b, Beijer et al. 2004, Farbry et al. 2001, Lee et al. 2004). Thus, it has been asserted that advertisements that are the most successful from a marketing perspective may be those that pose the greatest threat to driving behavior (WHO 2011). A more formal discussion on the relationship between distraction and specific infrastructure elements is provided in Chapter 3. The available studies all agree that elements of the actual roadway environment should be included as sources of distraction outside the vehicle. However, collectively, these sources essentially include any element outside the vehicle that focuses the driver’s attention away from the driving task. To mitigate such distractions, transportation agencies often have statutory authority to prohibit outdoor advertising that interferes with traffic safety. For example, a Minnesota statute prohibits “signs…which are of such intensity or brilliance as to cause glare or to impair the vision of the driver of any motor vehicle, or which otherwise interfere with any driver’s operation of a motor vehicle” (Minnesota Statutes §173.16). Many transportation agencies are also authorized to review (from a traffic safety perspective) the plans for new buildings and structures adjacent to roadways under their jurisdiction. Remedial measures for existing distractions include walls, fences, glare screens, and landscaping to limit views of activities that are not relevant to the driving task. Together, these definitions of OVDs, statutes, and measures potentially comprise a fairly wide range of distractors and interventions. Therefore, a key challenge for BTS-09 is directing the limited project resources toward the situations that are of greatest relevance to the needs of agencies. The definition of IRDs can first be narrowed to fixed objects, since the objective of BTS-09 is infrastructure related. This restriction excludes objects such as people, animals, moving cars, debris, weather, or natural features (e.g., recreational areas, waterfalls). Moreover, the ultimate goal of the Safety Frameworks developed for this project is to identify characteristics that are correlated to distraction in order to either mitigate the impacts of or develop policies that would

14 restrict or limit those infrastructure elements. As a result, the focus can be further narrowed to objects that reside within the roadway or on the roadside that a roadway agency is responsible for or is likely to be able to address through policy. Using the taxonomy developed by Horberry and Edquist (2009), this narrower definition of IRDs would include objects along the built roadway and some features in the natural and built environment but would not include situational entities. For the purpose of BTS-09, the following is the operational definition of an IRD: An infrastructure-related distraction is a glance away from the driving task related to a fixed object that is part of the roadway cross section, including the roadway itself, the shoulder, the median, and any right-of-way owned or operated by the corresponding roadway agency, or a fixed object for which a transportation agency is reasonably likely to be able to implement a policy addressing the distraction (e.g., wind turbines or monuments) or that is placed adjacent to the roadside specifically for the traveling public (e.g., billboards). 2.3.2 Surrogates for Infrastructure-Related Distraction Researchers have commonly defined distraction in terms of glances of a certain length away from the roadway task (e.g., 2 seconds). However, for some data sets, it is difficult to evaluate distraction using this definition. For instance, in NDS data sets, finding a sufficient sample of glances away from the driving task that last at least 2 seconds to demonstrate the safety impact of a particular distraction is challenging due to the difficultly of locating these activities among the often vast amount of data within these data sets. For instance, the SHRP2 NDS data contain more than 5 million trips with over 35 million vehicle miles recorded. Another challenge in relating distractions of a certain length to infrastructure features is that the amount of time that a driver is able to glance at a particular element is limited. For instance, assuming that a railroad crossbuck sign has approximately the same visibility as a static roadway sign (around 180 ft), at 40 mph a crossbuck sign is only visible to a driver for 3 seconds. As a result, it is difficult in any data set to identify long glances away from the driving task when the window of opportunity to view the sign is short. It is similarly difficult to identify longer glances in simulator data sets due to the short length of driving activity as well as the inability for simulators to fully replicate scenarios that result in driver distraction. As a result, identifying glances away from the driving task of 2 or more seconds is challenging for many infrastructure elements. Another challenge in assessing distractions related to infrastructure is that, in most cases, drivers glance at infrastructure features in the same location as they look when performing regular driving tasks. This makes it difficult to differentiate when a driver is glancing at the roadway in order to attend to the driving task (which may include glancing at the feature in question) or fixating on an infrastructure element to the detriment of the driving task. Due the challenges of using common visual distraction metrics, it may be necessary in some cases to utilize surrogate measures to assess the impact of infrastructure on distraction. Researchers have utilized a number of metrics that may be correlated to distraction. Lateral

15 control measures such as lane position, standard deviation of lane position, and lane excursions have been shown to have a relationship with distraction and have been used as surrogate measures (Green et al 2004, Ka et al. 2020). For example, Ranney et al (2011) evaluated the impact of various in-vehicle information systems and found increases in standard deviation of lane position compared to the baseline scenario when drivers were engaged in various tasks. Törnros and Bolling (2005) found an increase in deviation in lateral position when drivers engaged in dialing tasks. Liang and Lee (2010) evaluated driver behavior before and during distractions and found that both visual and cognitive distraction tasks impacted standard deviation of lane position, with visual distractions having the greatest impact. Young et al. (2013) evaluated driving errors using an instrumented vehicle and found that drivers engaged in distracting tasks were 24% more likely to have a lane excursion. Speed, standard deviation of speed, and acceleration/deceleration have also been found to be correlated with distraction and have also been used as surrogate measures. Egström et al. (2005) evaluated simulator and instrumented vehicle driving and found that visual demand led to reduced speeds. Törnros and Bolling (2005) evaluated the impact of dialing a cell phone on driver performance and found that driving speed decreased as drivers engaged in dialing tasks. Young et al. (2013) evaluated driving errors using an instrumented vehicle and found that drivers exceeded the posted speed limit 34% more frequently during distraction events than during baseline events and traveled too fast for turns 54% more frequently. Additionally, they found higher mean speeds and significantly higher standard deviations of speed when drivers were engaged in distractions and found that twice as many drivers accelerated too fast when distracted compared to the baseline. Other measures used to infer distraction include erratic acceleration and deceleration, number of fixations, erratic steering behavior, and deviations in steering wheel position (Ka et al. 2020, Green et al. 2004, Ranney et al. 2011). Liang and Lee (2010) also noted that cognitive distraction disrupts the allocation of visual attention to the driving scene and can cause drivers to concentrate their gaze in the center of the driving scene. This type of distraction is often measured by the horizontal and vertical standard deviation of gaze distribution. This finding suggests that a driver’s fixation on one location of the windshield may also be used to indicate distraction in relation to an infrastructure element. Based on the above discussion, the following surrogates may be the most relevant to identify the relationships between distraction and infrastructure elements. A number of these surrogates were tested in the various Safety Frameworks that were developed for this research (as described in Chapters 6 through 10). Potential surrogates include the following: • Glances at forward roadway scene (may suggest fixation on an infrastructure element in the forward roadway scene): o Percent road center (PRC) gaze: Percentage of time when gaze is directed to the forward roadway within the epoch (defined as the amount of time or distance within the study period of interest) o Number of glances away: Number of glances away from the forward roadway within an epoch.

16 • Glances toward specific infrastructure element (when infrastructure elements can be differentiated from forward roadway scene): o Number of glances: Total number of glances to each element. o Average glance time: Average duration of glances to the element. o Total glance time: Total eyes-off-road time associated with each element. o Horizontal gaze dispersion: Horizontal dispersion of gaze (in degrees). o Vertical gaze dispersion: Vertical dispersion of gaze (in degrees). • Lateral control (which indicates cognitive workload): o Standard deviation of lane position: Indicates variation within lane. o Lane excursions: Number of times vehicle crosses lane line. • Speed control (which indicates cognitive workload): o Mean speed: Average speed for vehicle within an epoch. o Standard deviation of speed: Indicates variations in speed. o Acceleration/deceleration: Measure of acceleration or deceleration that exceeds threshold of normal behavior. 2.4 Summary This chapter began by highlighting definitions for distraction in relation to driving. The chapter then described common measures utilized to assess distraction in research. These included visual distractions of greater than 2 seconds for simulator or naturistic driving studies and crash reports that have a source of distraction code (S2) of 06 External (to vehicle/nonmotorist area) based on the MMUCC-5 data model. The chapter concluded by using past research to develop a detailed working definition of IRD and then listed potential surrogate measures of distraction. The use of this operational definition and these potential surrogate measures in the development of the Safety Frameworks is described in Chapter 4. Before these Safety Frameworks were developed, however, a literature review of studies on known IRDs was conducted. In the next chapter, the body of knowledge related to infrastructure elements and distraction is summarized.