Most deaths among young people between 15 and 29 years old are caused by traffic accidents. Traffic accidents are also one of the leading causes of death, injury and disability in other age groups. In Hong Kong, about 20,000 people are killed or injured in traffic accidents every year; across the border in mainland China, the picture is grimmer with some 260,000 people losing their lives annually — or an average of more than 700 a day. Is there a way of identifying risky drivers before they cause an accident instead of identifying them after the fact? It seems a specially designed video-game-like simulated driving task can.

As part of its Decade of Action for Road Safety to reduce the number of road fatalities by 2020, the United Nations is promoting regulatory measures that all governments can consider to reduce deaths from 'external' risk factors, such as reducing and enforcing speed limits and implementing certain standards for vehicular equipment. However, an international team of researchers decided to tackle the arguably harder-to-control 'internal' risk factors of drivers' risk taking and risky decision making, which previous studies have indicated as also contributing to traffic accidents and violations.

The team, which included Dr Andy S.K. Cheng, Associate Professor at PolyU's Department of Rehabilitation Sciences, used a desktop computer, a computer monitor, and a video-game steering wheel and pedals to create a driving simulator to measure drivers' decision making and behaviour when faced with 5 different virtual-reality driving scenarios. Since the subjects were Chinese drivers on the mainland, the scenarios were designed for driving on the right-hand side of the road. The scenarios were driving straight on a straight road, driving straight on a straight road with an adjoining side road on the right (effectively a T junction or intersection), driving straight on at a crossroads or X intersection, making a right turn at a crossroads, and making a left turn at a crossroads.

Eighty-four licenced drivers who met various driving-experience criteria and were willing to furnish their record of traffic violations and crashes over the 3 previous years gave their consent to take part in the study. They were asked to complete 3 existing questionnaires and tests that are widely used to measure drivers' deliberately dangerous and unintended erroneous driving behaviour, anger trait in general, and tendency to take risks, respectively. The results of these would be used to help evaluate the effectiveness of the research team's simulated driving task as a new way to detect risky drivers. Each driver then underwent a 15-min driving training session on the simulator to familiarise him or her with it. For the simulated driving task, each driver was asked to complete an identical simulated urban route in the shortest possible time but without violating any traffic regulations or causing any accidents. They were told that the quicker they took to complete the route, the more reward money they would receive according to a formula (with RMB20 as the maximum amount). However, they were also told that each traffic offence they committed and each accident or crash they caused would incur a monetary penalty of different but much smaller amounts.

In the simulated driving task, each of the driving scenarios appeared twice using the same road layout: the first time as a 'baseline measurement' scene in which the road, junction or crossroads was clear of other road users; and the second time as a 'dilemma' scene in which the driver encountered other (simulated) road users who could get in the way, forcing the driver to decide whether to take a risk and keep going (a "go decision") or be cautious by waiting for the other users to stop or finish their manoeuvres first (a "no-go decision"). Like in a video game, if a driver crashed or caused an accident, his or her car was reset onto the road so he or she could continue driving. Several driving-related parameters were recorded or calculated by the simulator's equipment and software during each driver's task.

After the simulated driving tasks were finished, the drivers' data were analysed. The research team divided the drivers into high risk or low risk depending on how many go decisions the drivers made during their task. They noted the high-risk drivers had a significantly higher number of real-world traffic violations compared with the low-risk drivers, and the high-risk drivers also had significantly more violations and crashes during their task.

The high-risk drivers also completed their respective task significantly quicker than their low-risk peers did. The high-risk group also scored significantly higher on the existing questionnaires and tests for deliberately dangerous driving behaviour, anger trait in general, and risk tendency. The team found that the risk taken during the simulated driving task seemed to be the result of intended risk taking by the drivers rather than the result of their unintended erroneous driving or anger trait.

The high-risk drivers also kept a significantly shorter minimum distance and had a significantly shorter time-to-collision to approaching road users (that is, the time the car took to hit the vehicle in front if it braked suddenly). The high-risk group pressed the accelerator pedal to a greater depth and the brake pedal to a shallower depth than did their low-risk peers. The high-risk drivers drove significantly faster and with a smaller variance of speed than their low-risk counterparts did. They also moved the steering wheel significantly more quickly and to significantly larger angles, and they performed significantly more steering-wheel corrective movements than did the low-risk group. The differences between the groups were larger during the dilemma scenes than during the baseline scenes.

Notably, the speed of the driver's car, the depth pressed on the accelerator, and how quickly the driver turned the steering wheel were statistically distinguishable between the high-risk and low-risk groups in all scenes and all scenarios, which point to their potential as possibly reliable measures of risk taking in real-world driving.