Should an autonomous car injure a pedestrian to save your life?

A study shows that autonomous vehicles could be programmed to make ethical decisions like those of a human.

A recent study from the Institute of Cognitive Science at the University of Osnabrück (Germany) has revealed that the decisions we make behind the wheel are not as complex as was believed, so they could be digitally programmed into autonomous vehicles through different algorithms. Braking, turning left or swerving around an obstacle are manoeuvres that autonomous systems already perform, but what would happen if the vehicle had to choose between hitting a pedestrian or crashing?

105 participants in a study on decision-making behind the wheel

The study, entitled “use of virtual reality to evaluate ethical decisions in road traffic scenarios”, was published in the Swiss journal Frontiers in Behavioural Neuroscience, where researchers from around the world publish studies on human and animal behaviour.

In the research, 105 participants were placed in a virtual reality scenario where they had to drive a vehicle on a foggy day. Along the route, the drivers encountered several obstacles and had to choose between hitting people, animals and objects with a virtual car.

Most drivers make similar decisions

The results of the study showed that decision-making is similar among the participants, especially in scenarios where there was the option of avoiding any obstacle. This human ethics showed that most drivers act in the same way, so the decisions could be applied through algorithms to autonomous driving systems that could make decisions by recognising the obstacles the vehicle encounters while driving.

However, Professor Gordon Pipa, the lead author of the study, warns of the dangers of this method due to the probability of failure. “Should autonomous systems fully imitate human decisions? Even the negative ones? And if things go wrong, whose fault is it?” the professor said.

In an unavoidable accident, the autonomous vehicle would have to decide for us

Developers of autonomous systems and specialists in psychology are investigating the moral dilemmas an autonomous vehicle would have to face in an unavoidable accident, and whether the car itself would have the capacity to make a decision similar to the one a human being would make in situations where there is no time to react. For example, imagine the car reaches a traffic light where a child and an adult are crossing illegally. The car has no time to brake and must change direction, inevitably hitting one of the two individuals. Who would it choose?

Save the passenger or the pedestrian? MIT’s test for resolving moral dilemmas

The prestigious Massachusetts Institute of Technology (MIT) developed last year a project called “Moral Machine”, a test in which users must choose between moral dilemmas posed from the perspective of an autonomous vehicle.

If you want to take control of an autonomous car that is going to suffer an unavoidable accident and decide between questions such as the following – which is more valuable, the lives of two children or of three adults? Would you save a driver before two pedestrians? – you can do so by clicking on the following image. After taking the test, you can consult the results of other participants and see how similar your answers are to theirs.

The great challenge of artificial intelligence: autonomous systems that take ethical issues into account

For now, artificial intelligence has not been applied in cars, but large companies such as Apple, Google, Nissan, or Tesla, are already developing systems that, in addition to cameras, sensors, GPS and radars, will have the capacity to make decisions.