Effectiveness of risk awareness perception training in dynamic simulator scenarios involving salient distractors

https://doi.org/10.1016/j.trf.2021.06.009Get rights and content

Highlights

  • A driving simulator and eye tracker were used measure latent hazard anticipation.

  • RAPT-trained drivers showed better hazard anticipation than Placebo-trained drivers.

  • The effect of RAPT persisted even in the presence of dynamic distractors.

  • The results suggest added benefits of RAPT beyond tactical hazard anticipation.

Abstract

The Risk Awareness Perception Training (RAPT) has been shown to improve latent hazard anticipation in young drivers. However, previous evaluation scenarios in a driving simulator often lacked either dynamic road environment features or control for such variations. The current study investigated whether the effectiveness of RAPT persists even in the presence of dynamic and salient distractors. Twenty RAPT-trained drivers and twenty-one Placebo-trained young drivers (aged 18–21) drove through eight simulated driving scenarios with latent hazards. A pedestrian avatar served as a distractor and was placed across from the latent hazard location. In half of the scenarios, the pedestrian remained static while in the other half the pedestrian started to move, without potential interference with the driver’s travelling path, as the drivers approached the latent hazard. Consistent with previous research, RAPT-trained drivers demonstrated better latent hazard anticipation performance than Placebo-trained drivers regardless of dynamic movement of the pedestrian avatar. Additionally, RAPT-trained drivers adopted wider scanning patterns and fixated more frequently on both the latent hazard and the pedestrian compared to Placebo-trained drivers. The results imply that RAPT may protect drivers from being distracted by dynamic stimuli and allow them to scan safety–critical areas containing latent hazards. Furthermore, RAPT may not only improve tactical hazard anticipation skills, but also modal hazard anticipation skills in young drivers.

Introduction

Although the number of traffic fatalities per mile driven has steadily declined over the past four decades (NHTSA, 2019), young drivers are still overrepresented in fatal crashes, accounting for 8.9% of fatal crashes in 2016, despite possessing only 5.4% of driver’s licenses in the United States. Contrary to a common notion that this overrepresentation in vehicular crashes stems from risky driving habits of young drivers, research has started documenting that attentional and cognitive failures while driving appear responsible for crashes in young drivers. For example, an analysis of 2128 non-fatal accident reports involving young drivers aged 16–19 (McKnight & McKnight, 2003) showed that failures in attention and failure to adequately search the roadway contributed to 23% and 42.7% of the accidents, respectively, while deliberate risky behaviors only accounted for approximately 5% of accidents, patterns that have been found in different young driver samples (Braitman, Kirley, McCartt, & Chaudhary, 2008, Curry et al., 2011). This seminal work highlights two important points. First, high crash rates involving young drivers are due to young drivers being clueless but not necessarily careless. Second, a large percentage of these crashes involving young drivers are caused by attentional or cognitive failures which are critical for young drivers’ road safety (Treat et al., 1979) and amenable to training.

More recent research in surface transportation human factors has identified three higher cognitive skills that are critical for road safety in young drivers –hazard anticipation (Fisher et al., 2007, Unverricht, Samuel, & Yamani, 2018), hazard mitigation (Muttart, Fisher, & Pollatsek, 2014, Muttart et al., 2013), and attention maintenance (Chan, Pradhan, Pollatsek, Knodler, & Fisher, 2010, Yamani, Bıçaksız, Palmer, Hatfield, & Samuel, 2018). Specific training programs that aim to train each of the three skills have been developed and evaluated such as RAPT (Risk Awareness and Perception Training; Pradhan et al., 2009) for hazard anticipation, FOCAL (FOrward Concentration and Attention Learning; Pradhan et al., 2011) for attention maintenance and ACT (Anticipate-control-terminate; Muttart et al., 2019) for hazard mitigation. Among the three, latent hazard anticipation is the skill that has been the most extensively studied and targeted for training program development and evaluation (Fisher et al., 2017, Fisher et al., 2007, Pradhan et al., 2005, Pradhan et al., 2009, Unverricht, Samuel, & Yamani, 2018). More specifically, latent hazard anticipation is the ability to anticipate hazards that exist on the forward roadway but have not yet materialized (Fisher et al., 2007, Pradhan et al., 2005). An example of a latent hazard is a truck parked in front of a pedestrian crosswalk such that the entrance to the crosswalk is obscured to approaching drivers. Consequently, a pedestrian that is about to step out into the crosswalk may be hidden by the truck. A safe driver, anticipating this latent hazard, would look towards the front of the truck as they approached the crosswalk to ensure that no one was in the entrance of the crosswalk behind the truck. Other examples of latent hazards include vehicles blocking the view of a bike lane or an overgrown hedge obscuring the entrance of a pedestrian crosswalk. Latent hazard anticipation performance is measured as the proportion of latent hazard scenarios where the driver correctly fixates on the location of the latent hazard, measured via eye-tracking (Unverricht et al., 2018). Using this measure of latent hazard anticipation, prior studies have demonstrated that young, novice drivers are worse at anticipating hazards than older, more experienced drivers (Pradhan et al., 2005, Taylor et al., 2011).

The question of how we can improve latent hazard anticipation is a critical one that goes beyond driver safety; 2019 had the highest number of pedestrian fatalities since 1988 (GHSA, 2020). A detailed analysis of vehicle crashes that resulted in a pedestrian fatality revealed that in 17% of cases, the responsible driver reported obscured view of the pedestrian (Jermakian & Zuby, 2011), highlighting the need to accelerate and instill latent hazard anticipation skills in young drivers for the safety of all road users. Researchers have developed several different training programs designed to improve latent hazard anticipation skills in young drivers (Borowsky et al., 2012, McKenna, Horswill, & Alexander, 2006, Meir, Borowsky, & Oron-Gilad, 2014). One such training program that has been developed and extensively evaluated is the Risk Awareness and Perception Training (RAPT; Pradhan et al., 2009, Unverricht, Samuel, & Yamani, 2018).

RAPT is a computer-based program that trains drivers to correctly anticipate latent hazards in a variety of driving hazard scenarios with specific feedback and practice. To facilitate learning, in the most recent iteration (RAPT-3), trainees are presented with an exocentric (top-down) view of the scenario that provides context and information on where hazards might be hidden. After being shown the exocentric view of the scenario, trainees are shown a series of static egocentric images (from the driver’s point of view), approaching the scenario. The trainee is instructed to click on areas of the photos where they think they should look for hazards if they themselves were driving through the scenario. If the trainee fails to identify the hazard, the trial repeats until the trainee correctly locates the hazard, or a maximum of three times. This strategy of providing both egocentric and exocentric views of hazard situations has been shown to be more effective at improving latent hazard anticipation performance than just providing trainees with only one viewpoint (Unverricht et al., 2018). The training program provides feedback to users and encourages trainees to learn from their mistakes and analyze the hazard situations carefully.

RAPT has been shown to improve latent hazard anticipation performance in novice teenage (16–18) and young drivers (18–21), evaluated both on the road and in a driving simulator (Fisher et al., 2007, Pradhan et al., 2009). For example, Pradhan and colleagues (2009) demonstrated that RAPT-trained drivers had 28.8 percentage point improvement in anticipating latent hazards on the road compared to untrained drivers. The measurable benefits of RAPT have been shown to persist up to six months after the initial training exposure, demonstrating potential long-term retention of the improved hazard anticipation for trainees (Taylor et al., 2011).

Studies have shown promise that improvements in latent hazard anticipation performance via RAPT can translate into safer driving on the road. In a large-scale naturalistic evaluation study, 5,251 young drivers in California (aged 16–18) were assigned to either RAPT (n = 2,663) or a control training program (n = 2,588) after passing the test to receive their driver’s license (Thomas et al., 2016). Using crash reports, in the year following training, male RAPT participants demonstrated 23.7% lower crash rate than those who completed the control training. This effect was not seen in female trainees, perhaps because teenage male drivers at the baseline are more likely to be involved in crashes (NHTSA, 2019) than teenage female drivers, therefore having more room for improvement. Although there is a possible gender difference on the effectiveness of RAPT, findings from Thomas and colleagues (2016) still show promise that improved latent hazard anticipation performance can translate into safer driving in young drivers.

To further predict and control the drivers’ scanning strategies, a theoretical framework involving bottom-up and top-down processes developed in Cognitive Psychology may lend important insights to analyze what factors influence attention distribution in young drivers. Latent hazard anticipation is a top-down, higher cognitive process, because the target (hazard) has not yet physically materialized, requiring the driver integrate context, cues, and prior experience in order to successfully anticipate the hazard. These actions are signatures of top-down processing (Cavanagh, 1991). However, the real-world driving environment has dynamic and salient stimuli that can attract attention away from safe driving in a bottom-up manner, which in turn could mitigate the effectiveness of RAPT in young drivers.

When on the roadway, drivers must engage in both top-down and bottom-up processing. Top-down processing is knowledge- or experience-driven (e.g. searching for a latent hazard) whereas bottom-up processing is stimulus-driven (e.g. bright colors, luminance) (Yantis, 1998). Filtering of irrelevant stimuli, even when attracting attention in a bottom-up manner, is critical for safe driving. For example, when viewing videos of driving scenes of developing hazards, participants were worse at predicting what would happen next in the scene when the video clips contained extraneous irrelevant, yet potentially hazardous, elements compared to when just one developing hazard was portrayed, illustrating the difficulty that drivers may experience when faced with complicated driving situations (Muela et al., 2021). Similarly, movement on the roadway can attract drivers’ attention in a bottom-up manner. Participants were more likely to fixate on objects that were moving in the periphery than those that were static in the periphery when watching video clips of driving scenes (Underwood et al., 2003).

Dynamic billboards also incorporate the elements of movement and can attract drivers’ attention away from the roadway. In an on-road study, the number of off-road glances and the duration of off-road glances increased when participants passed a dynamic billboard compared to when participants were driving in areas without any billboard (static or dynamic) (Belyusar et al., 2016). In the same study, drivers glanced at the dynamic billboard more frequently when it was changing messages than when it was static. This finding suggests that the movement onset of a message can draw attention more frequently than the illumination of the billboard without such movement. The movement onset has already been demonstrated to attract attention in the standard visual search task (Abrams & Christ, 2003; also see Wolfe & Horowitz, 2017) and may be a critical bottom-up feature that can detract from latent hazard anticipation, a form of top-down processing, critical for safe driving.

Although RAPT has been shown to improve latent hazard anticipation performance, it is still unclear how persistent the effects of RAPT would be in the presence of extraneous, dynamic stimuli. The effectiveness of RAPT has been evaluated on the road in realistic environments, but these on-road evaluations lack control of extraneous, dynamic stimuli (Fisher et al., 2007). The current study examined whether trained young drivers were able to anticipate latent hazards, even in the vicinity of dynamic, driving irrelevant stimuli. Movement onset has been shown to attract attention away from the forward roadway as seen in a dynamic billboard study (Belyusar et al., 2016). Therefore, the current study will use movement onset of a pedestrian avatar as the task-irrelevant feature in the hazard scenario. Because RAPT has been shown to be effective at improving latent hazard anticipation performance in uncontrolled, dynamic scenarios (Fisher et al., 2007), we hypothesized that RAPT-trained drivers would anticipate hazards more accurately than placebo-trained drivers regardless of the presence or absence of movement of a task-irrelevant object in the driving environment.

Section snippets

Participants

Forty-one young adults, aged 18–21, enrolled in a southeastern university completed the study; 20 participants were assigned to the RAPT group (six males, mean age = 18.90 years, SD = 0.91; mean months since licensure = 26.3, SD = 14.40) and 21 participants were assigned to the Placebo group (three males, mean age = 18.81 years, SD = 0.75; mean months since licensure = 28.0, SD = 9.76). There were no measurable differences between the two groups in terms of age, t(39) = 0.35, B10 = 1/3.12, or

Results

Eye tracking failures resulted in data loss in 2.7% of trials, with no significant difference in proportion of data loss between groups (p = .331). All analyses were conducted on participants’ mean score by Pedestrian Type (moving or static pedestrian avatar). Table 1 displays means and standard deviations of latent hazard anticipation and pedestrian fixation proportion.

Replicating previous studies, data provided decisive evidence that RAPT-trained participants correctly anticipated more latent

Discussion

The purpose of this study was to examine the potential effect of an additional dynamic element on latent hazard anticipation performance in young RAPT-trained and Placebo-trained drivers using a driving simulator. Consistent with previous research, latent hazard anticipation performance for RAPT-trained drivers was better than Placebo-trained drivers (Fisher et al., 2007, Pradhan et al., 2009, Yamani, Bıçaksız, Palmer, Hatfield, & Samuel, 2018). This study further demonstrated that the RAPT

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

CRediT authorship contribution statement

Sarah Yahoodik: Conceptualization, Methodology, Formal analysis, Investigation, Writing - original draft, Writing - review & editing. Yusuke Yamani: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing, Supervision.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (45)

  • I. Muela et al.

    Visual attention in realistic driving situations: Attentional capture and Hazard Prediction

    Applied Ergonomics

    (2021)
  • J.N. Rouder et al.

    Default Bayes factors for ANOVA designs

    Journal of Mathematical Psychology

    (2012)
  • G. Underwood et al.

    Driving experience, attentional focusing, and the recall of recently inspected events

    Transportation Research Part F: Traffic Psychology and Behaviour

    (2003)
  • Y. Yamani et al.

    Impact of information bandwidth of in-vehicle technologies on drivers’ attention maintenance performance: A driving simulator study

    Transportation research part F: Traffic psychology and behaviour

    (2018)
  • R.A. Abrams et al.

    Motion onset captures attention

    Psychological Science

    (2003)
  • Cavanagh, P. (1991). What’s up in top-down processing. In A. Gorea (ed.) Representations of vision: Trends and tacit...
  • D.E. Crundall et al.

    Effects of experience and processing demands on visual information acquisition in drivers

    Ergonomics

    (1998)
  • D. Crundall et al.

    Attending to the peripheral world while driving

    Applied Cognitive Psychology: The Official Journal of the Society for Applied Research in Memory and Cognition

    (2002)
  • D.L. Fisher et al.

    Handbook of teen and novice drivers: Research, practice, policy and directions

    (2017)
  • D.L. Fisher et al.

    Empirical evaluation of hazard anticipation behaviors in the field and on driving simulator using eye tracker

    Transportation Research Record

    (2007)
  • Governors Highway Safety Association (2020). Pedestrian Traffic Fatalities by State, 2019 Preliminary Data....
  • H. Jeffreys

    Theory of probability

    (1961)
  • View full text