Driverless cars ‘could choose who gets hit in accidents’
Lawyers warn that autonomous vehicles may be forced to settle moral dilemmas for themselves, such as no-win situations
Driverless cars could be programmed with their own morality about who to hit in a collision.
A report by the Faculty of Advocates – an independent Scottish body of lawyers – suggests it may be near-impossible to determine the behaviour that causes accidents involving autonomous vehicles. This is because while current technology is based on set algorithms, rules and processes, manufacturers are seeking to develop ‘neural networks’ that can make their own autonomous decisions.
Responding to a joint consultation by the Scottish Law Commission (SLC) and the Law Commission of England and Wales, which are working to prepare laws for future self-driving vehicles, the faculty said: “It is a feature of such systems [neural networks] that their internal ‘reasoning’ processes tend to be opaque and impenetrable (what is known as the ‘black box’ phenomenon) – the programmers may be unable to explain how they achieve their outcomes.”
One particular sticking point is the car’s morality – usually presented as a question of who an autonomous vehicle should try to spare when a collision is unavoidable. Examples include swerving out of the way of a child only to hit an elderly person, or favouring a professional over a criminal.
The report said: “Persons generally are entitled to expect that a self-driving vehicle will not collide with and injure them. However, in reality the situation is much more nuanced.”
It was even suggested that people may be able to choose their car’s ethical preferences while also deciding on the paint colour and interior trim when buying it.
The UK government says fully driverless cars will be on use on the roads within two years, meaning the answers to questions such as this are vital.
SLC commissioner Caroline Drummond said the consultation, which closed yesterday, had many responses.
“The advent of driverless cars raises an enormous range of aspects, such as jay-walking, which we do fairly freely here but isn’t allowed in other countries. It certainly won’t work with autonomous vehicles.
“From an engineering perspective, from an AI perspective and from a social perspective, there are a lot of areas here that really need to be explored,” she said.
