Skip links

Ethical Challenges in Self-Driving Cars

Blog. How will self-driving cars change transportation and our cities, who is to blame in an accident, are they creepy or not. The newspaper Information hosted an exciting event at Science Day, 25.04.2016. Below are selected highlights.

Vi er primært på niveau 1 - nogle få på niveau 2. Slide fra Mette Møller
Most cars are in level 1, few in 2 in the automization process

Mette Møller, DTU: 80-90% of all accidents are caused by humans. Our brains are not designed to assess distances at a fast speed. And then there are challenges as alcohol. So, we are facing massive security improvement possibilities. The big challenge is the transition, and we are only just at the beginning. Research must answer a series of questions like ‘what will human drivers do, when they don’t drive the car’, ‘is it necessary to differentiate between drivers’? One conclusion is that the human factor cannot be completely eliminated. It’s also humans, who must program the cars and ensure that human errors are eliminated.

Malene Freudendal Petersen, Roskilde University: Self-driving cars are amazing, but they will not solve such problems as pollution, traffic- and parking chaos. Self-driving cars are not just a technological quick fix.

Jakob Holtermann, University of Copenhagen. Who is responsible for an accident in a self-driving car? Perhaps the concept of guilt and responsibility is not so useful. Maybe we should make a NESS test – necessary element of a sufficient set. The fact that I press the switch may not be enough to light up a bulb. For the bulb to work, there must be electricity and so on. This approach is particularly useful in accidents in complex man-made safety systems such as airports and hospitals. The SAS airplane-accident in Milan 2001, for example, there were many Ness conditions: fog, lack of ground radar, mistakes with traffic controller, ambiguous light signal, misleading map of the airport etc. It is hard to point to one guilty person or entity. And who do you threaten with punishment, when it comes to self-driving cars? The driver, employees, business management, authorities? The threat of punishment can also cause people to hide dangerous situations. So, we have to think about prevention – also in the construction of systems.

Helle Spindler, Aarhus University. The question is how perfect we can make the cars. What is the perfect algorithm – and do we take an individual or societal perspective.

paretoMartin Mose Bentzen, DTU. Can we build ethical robots? We can chose to use e.g. the Pareto-optimization, which means that the actions we give the robot should make the situation better for as many people as possible. But who should decide how the self-driving cars should behave? The producers, the politicians or the users – who can put it in ‘egoist-mode’. Martin Mose believes that producers should not determine it themselves.

'sød'bil
Robotic cars often have cute faces

Gunhild Borggren, , Copenhagen Universitet. People prefer interacting with robots that look like people. Hence, many robots are equipped with a face. Looking at cars, we see many of them having a front that look like a human face. But it can also be too much. A Japanese researcher has proven that there is a creepiness factor. This need to be examined it in relation to cars.

The debate, guided by the moderator Henrik Føhns, followed.

DataeEtichs asked what the panel thinks about all the data collection that follows the self-driving cars.

Martin Mose Bentzen says we should separate the two debates – the technical self-driving car and the collection of data. But Malene Freudendal Petersen points out that it is super important – perhaps the most important to earn trust – and that we definitely should not separate the two things.

Nikolaj Nohr-Rasmussen, site manager at Volvo: In principle, the car runs without collecting personal data. So technically it’s a different discussion – it is two parallel tracks. Self-driving cars should be designed so that they are not dependent on being online. So, in principle the cars are not collecting data. But we want to learn from all this. And then data collection is very important, but we can make the data anonymous, which we do.

Mette Møller adds: From a safety aspect, I can say that the more we know, the better.

Bo Ekman, the Danish Road Directorate: It is important that producers are not sitting on all the data, but that they share it with authorities to improve road safety.

The difference between the driverless car and your smartphone is that it has fatal consequences if your car does not work, but not with your smartphone. Therefore it does not go as fast with self-driving cars, explains Mogens Kjærgaard Møller from the Road Safety Council.