Worldwide, 1.35 million people die annually, and another 50 million are injured due to road traffic accidents [1]. Most of these accidents occur in middle and low-income countries. In Ecuador, a middle-income country, in 2018, 2,151 people died, and around 20,000 were injured due to road accidents [2]. One of the principal causes of these accidents is the human factor. In this factor, novice drivers are those who have a higher number of accidents compared to experienced ones [3]. Further details about the cognitive, behavioral, and attitudinal characteristics that define those drivers, can be seen in [4].

Candidate drivers can quickly acquire the elemental skills of controlling the vehicle and learning about traffic laws. This learning can be achieved with only 15 hours of driving [5]. However, it takes them much longer to develop perceptual and cognitive skills to move safely in a road environment [3]. Lack of experience may be associated with problems such as overestimation of their own abilities, underestimation of dangers, and longing for excitement and sensation [4]. Also, the human mood can affect young drivers, resulting in risky driving behaviors like transient violations (e.g., speeding) and risky exposure (e.g., driving tired) [6]. Among the skills acquired through driving experience, hazard perception is the most explored [7].

Hazard perception is the ability to read the road [8]. This perception can impact the driver behavior on the roads [9]. Novice drivers, in contrast to experienced drivers, can detect fewer hazards [10], may take longer to react [11], may skip areas with potential dangers [12], or may have difficulty processing them [13]. This subjectivity may lead them to choose inappropriate actions [14] that can lead to traffic accidents. It is difficult for the novice driver to detect the dangers since it is a multidimensional and complex cognitive process [15], which can only be acquired through age and experience. For example, drivers should mainly keep the vehicle on the road. Besides, they must identify the possible dangers that could lead them to a traffic accident. They should rule out low or impossible hazards and focus on the most dangerous ones. This process is continuous along the route and generally occurs within a short time. When speaking of the dangers, it is understood that these come from sources that the driver can see, regardless of whether or not they are aware of their effects [7]. On mountain roads, this process is more complicated due to their design and operation restrictions. Ideally, the novice driver should have a hazard perception close to the actual hazard in order to drive safer.

The actual hazard is the objective measure of the danger, that is, what can increase the frequency and severity of traffic accidents. There are several tools to measure the actual hazard. One of them is the methodology proposed by iRAP [16]. This tool allows to classify, using stars (1 = worst, 5 = best), the probability of an accident occurring, and how serious it can be for pedestrians, cyclists, motorcyclists and vehicle occupants. This star rating is calculated based on the particular characteristics of a site and from the results of previous investigations of traffic accidents. For example, a highway with low operating speeds is less dangerous to vehicle occupants than one with higher operating speeds. A place without a crosswalk is less dangerous to pedestrians than another one with a crosswalk. This actual danger should fit in the perception of the drivers, which is difficult since it is a subjective measure that depends on each driver.

The hazard perception can be measured using various tools (more details in [17]). Photos [18], [19], videos [7], [18], [20], simulators [21], or real trips [22] have been used. A survey or an interview was employed to collect data on these tools, and also, more details could be obtained about the drivers. Despite a large amount of research performed worldwide, very little has been done in Latin America where there is an increasing number of traffic accidents [1].

In this scenario, the purpose of this research is to analyze the hazard perception of novice drivers on mountain roads. The experiments performed consisted in showing short videos of several road environments to the participants. Thirty-seven novice drivers participated in the study. To present the results of those experiments, the remainder of this paper is organized as follows. The Materials and methods section describes details of the analyses: driver and site selection, the survey and data collection, and processing. Then, the Results section discusses the perceived and objective hazard, and its errors. Finally, the last section presents the main conclusions of the study.

This is an experimental investigation to assess hazard drivers' perception using videos of mountain roads. The data collection was conducted through online surveys due to restrictions by COVID-19 [23]. The surveys asked participants how dangerous the site was and why. After data collection, their answers were processed to get the study results.

This paper aims to analyze the hazard perception of novice drivers on mountain roads. Twelve road sites were presented in short videos to 37 novice drivers. The videos showed mountain roads from Ecuador and the United States of America. The hazard perception was compared to the objective hazard by iRAP star rating. After analyzing the results, the main conclusions are presented.

Novice drivers (men and women) do not correctly detect hazards in the most dangerous places (1 and 2 stars), as well as in the sites with a medium level of danger (3 stars). The perception improves in less dangerous places (4 stars). These trends were also found in the previous literature [18]. This can be explained using a previous study [19], where 90% of drivers, regardless of their experience, detect high threat objects placed on the side of the road. However, experts can identify a higher number of medium threat objects than novice drivers (55% of expert drivers versus 40% of non-expert drivers); while novices detect low threat objects (almost 20% of non-expert compared with none for the expert drivers). Therefore, proper object detection and prioritization could be achieved through training rather than years of experience. On the other hand, the errors between objective and subjective hazards were not statistically different between men and women. There was also no trend regarding the age of novice drivers. The mean errors for each star had a significant inverse relationship to the objective hazard. Errors are bigger in more dangerous locations and decrease as the hazard decreases. Several equations were calibrated, and these could help in future studies, for example by relating them to traffic accidents with these types of drivers.

Regarding the road attributes, novice drivers mainly detect hazards with horizontal or vertical traffic signs. A lower percentage of them also found hazards at the roadside, which are related to the crash severity. Many of them did not find any dangerous attributes in the sites which are impossible on this type of road. These results show that there is a need to improve the novice driver training process, including elements related to the driver and the road itself. Drivers indeed learn to detect hazards more effectively with increasing age and experience, but it is necessary to include formal education to try to accelerate that process.

This study has several limitations. First, it only had 37 novice drivers, yet the sample has a similar number of male and female participants. It was performed using videos and through on-line mode; however, through this method, resources can be saved and compared to real trips. Besides, this method is safer since the driver is not exposed to a crash accident during the collection of data. Despite these limitations, this research deepens the understanding of the relationship between the actual hazard and subjective hazard perception of novice drivers, who have a higher number of traffic accidents compared to other types of drivers. Also, the study focuses on mountain roads that have a high workload when driving. Also, it shows an experiment in the Ecuadorian context, which considers the influence of the local culture. Finally, these results help focus the resources of public and private institutions and achieve better training in the hazard perception for novice drivers. Better-trained drivers will be less likely to be involved in a traffic accident, thus avoiding fatalities on the roads.

The authors acknowledge the support of the National Secretariat of Higher Education, Science, Technology and Innovation (SENESCYT) and Universidad Técnica Particular de Loja from the Republic of Ecuador