Predictive Assessment of Driver Errors Using Human Error Template (HET)

Abstract Background an aims: Worldwide statistics declare that 1.2 million people are killed as a result of a road traffic accident (RTA). Besides, 20-50 million are injured every year. Traffic accidents account for leading causes 1.2% of deaths and 23% of injuries. In Iran, 16,000 people are annually cut short in traffic accidents. A great number of people suffer from non-fatal injuries/disabilities, which is estimated not to be less than 335,000. Statistics show that road traffic mortality in Iran has been higher than the global average, which can dramatically increase direct (medical costs, caring disabled people, broken cars, and etc.) and indirect (PTSD, traffic jam, depression in families, losing powers (permanently or temporarily) and etc.) costs. The magnitude of RATs in Iran can be better understood if it is mentioned that accidents are in the position of first place of the ranking of lost years of life. This caveat calls for a focus for urgency in RATs control strategies as a main focal point to improve public health in Iran. Nowadays, a triple-causal approach (vehicle, environment, and human) was accepted in traffic incidents. During recent years, vehicle and environment (as causal factors of traffic incidents) have been greatly improved. In contrast, human behavior has still remained as the most frequent contributing factor (up to 94%). This factor is being the most important and critical factor within any system, including traffic system. In a dangerous traffic situation, which is mostly no room to commit any error, accident prevention remains on drivers' abilities and skills. It seems human factors can play a special role to control human behavior. Therefore, it worth to provide a human factor view on human behavior while driving. During driving, the driver has to collect a large amount of information at any given time and put them in the process of continuous decision making. If any functions of the driver's decision-making process cannot be at the optimum level, human error may lead to catastrophic consequences. Driver errors are usually outcomes of mismatch between driving demands and driver abilities, especially psychologic ones. It is better to say human error is mostly the result of defects or improper function of mental information processing, which is familiar to people as forgetfulness, inattention, carelessness, negligence, recklessness, and etc. A group of methodologies which has been introduced in human factors are able to predict cognitive drivers’ errors. Despite the importance of role of human error in traffic accidents, Human error identification methods (HEIs) have been mostly used in high risk environments such as airline industry. Yet, there is few published studies on identification of driver’s errors and transportation industry. The purpose of this study is to predict driver errors using Human Error Template (HET) method meanwhile a real driving task. Methods: This case study was carried out to identify and predict the drivers' error in a specific real driving scenario. At first, the scenario was designed based on agreement of research team members. To fulfill the scenario, an Iranian car (PARS Peugeot) was used which had successfully passed the technical test (with regard to lights, mirrors, glasses, safety belts, horn, oil, outlet emissions, side slides, shock absorbers, brakes, steering, suspension system lever). The driver was a healthy 42 years old man with 20 years driving experience as his career. A direct route with proper traffic signs was considered on a two-way street with mild traffic flow on a sunny day. The scenario was as: “the driver departs from the park and moves in a pre-determined direction. A few moments later, he speeds up to overtakes the front car. Then, he simply continues driving. By getting the destination, he exits the path and park the car”. Since driving consists of several sub-tasks which should be performed simultaneously, a list of sub-tasks which are required to complete the scenario was provided by interviews, as well as, direct observation. The tasks were analyzed based on Hierarchical Task Analysis (HTA). Then, the tasks list was used as the input for the HET technique. This technique identifies and classifies external errors (EEMs) in the form of human error detection methods (HEIs). The technique considers EEMs in 12 patterns that are described as follows: 1. Fail to execute, e.g. the driver cannot correctly get the clutch 2. Incomplete task execution, e.g. leaving handbrake in middle position 3. Task execution in a wrong direction, e.g. pressing the gear lever in a wrong direction 4. Wrong task execution, e.g. suddenly change direction 5. Task repeated, e.g. placing the lever twice 6. Task execution on the wrong interface element, e.g. pressing a pedal instead of the others. 7. Task execution too early, e.g. turn on the guide much earlier than the redirection. 8. Task execution too late, e.g. delay in coming back to appropriate line and staying in the overtaking line. 9. Too much task execution, e.g. continuous gear changing 10. Too little task execution, e.g. few frequencies of using brakes 11. Misread information, e.g. misreading information on speed gauge 12. Other. Next, errors types and potential consequences were described. Results: Hierarchical task analysis (HTA) showed that for implementation of the scenario, three main tasks and nine sub-tasks were needed. Among 101 errors detected by HET, 16% were considered as unacceptable errors. Also, the most common types of errors included: “Task execution incomplete” (25.74%), “Task executed too late” (21.78%) and “Fail to execute” (13.86%), respectively. The results from the distribution of human errors in the main tasks and sub-tasks indicate that the main task “scrolling the route” accounts for 48% of all errors. Distribution of this ratio for “acceleration changing”, “line changing” and “distance adjusting” was 49%, 33% and 18%, respectively. Rate of errors were 33% in task of “moving” of which 79% were related to the “starting point of the movement” sub-task, 18% to the “alarms control” sub-task, and 3% to the “fasten seat belt” sub-task. Also, 19% of errors were found in the “parking” task, which 79% were related to “stopping” sub-task, 16% related to “shut downing”, and 5% related to “unlocking seat belt”. The distribution of catastrophic errors for the “scrolling the route”, “moving” and “parking” were found 4, 11, and 1 error, respectively. These were considered as "unacceptable" errors due to the high rate of incident occurrence (for example, suddenly line changing, lack of control of one of the front or rear line, etc.). Conclusion: The majority of identified type of errors in this study were “Fail to execute”. Therefore, to reduce this type of error, it is necessary to use corrective actions, such as periodic retraining drivers, equipping vehicles with visual and audible markers and alarms in relation to incomplete tasks. These actions can have a beneficial effect on reducing the severity and chances of occurrence of errors. Human Error Template (HET) is a comprehensive method that has been used widely in the aviation industry. As previous studies, the results of this study showed HET is capable to identify and classify driving errors. Using the technique in traffic domain can provide a great opportunity to predict drivers’ error. Consequently, there will be a great hope to control them before-the-fact. In spite of the importance of identifying drivers’ error, a few studies have been published. It seems the multi-sub-tasks nature of driving has caused the researchers to avoid involving in HTA for driving. To have a better understand of control strategy in traffic domain, however, it is strongly recommended to apply human error techniques for driving situation. Human factors