> Automated driving is necessary if the systems cause fewer accidents than human drivers.
True, but sounds like autonomous vehicles were a bitter pill to swallow for less accidents.
> In the case of unavoidable accidents, classification of drivers is prohibited. This includes determining fault by age, sex, or physical and mental constitution.
Does that the imply that we should only be in favour of auotonomous vehicles if they are always better than the worst drivers?
EDIT: I probably misunderstood this one. After reading the passage in the origal report I think it means that the computer is not allowed (for example) to prefer to save a child over an adult.
> We must be able to determine the responsible party in an accident at any time: human or computer.
Desirable, but good luck with that one.
> Drivers must have proper documentation.
Not sure what that means.
> The driver should be able to decide who owns his vehicle data.
Why not simply demand that the driver owns the data. If we leave the decision to the driver we will know how it ends up. They will assign away all their rights for a free cookie.
Most importantly the article (didn't read the report) skips one of the most important and difficult ethical questions. How the balance the life of other with the physical integrity of those in the car, especially in the light of unavoidable false positives.
The "proper documentation" part is about hybrid systems where the driver can be given control/responsibility in some situations. These situations must be clearly documented and should also be internationally standardized.
They also specifically warn and advise to prevent data collection practices currently used by Search Engine and Social Media Companies.
> Why not simply demand that the driver owns the data. If we leave the decision to the driver we will know how it ends up. They will assign away all their rights for a free cookie.
The translation of the article isn't great. With "owning" data they mean to determine who has access to the data. The driver has to be able to deny companies using their data. If this is implemented it would be a good step, Tesla currently doesn't allow that.
- If it can be shown that self-driving systems will lead to fewer accidents, the government will have an obligation to allow them–even if some accidents may continue to happen.
- The focus should be on avoiding those situations where technology has to make morally ambiguous decisions
- Where such life-vs-life decisions do happen, it would be immoral to assign different values to the individual lives on the basis of sex, age or other characteristics.
- In principle, it would be possible to allow decisions trying to minimise the lives lost/put at risk, i. e. sacrificing one life to save two.
- But: they also deem it immoral to sacrifices the life of people who did not participate in the creation of the traffic risk themselves.(1) This means, for example, that system may not "sacrifice" one pedestrian for two (or even one) passengers of the car.
- Liability will move from the owner/operator to the producers of cars
- The technology, and the decisions it implements, must be communicated to the public
- It's unclear if traffic should be directed by some central authority. Such centralisation would need to find a way to operate in a way that makes it impossible to create profiles of individuals' movements
- The car's owner must always agree to the sharing and usage of their data. This requires not just the theoretically choice, but the real possibility to say no, and realistic options to do so without undue consequences.
- It must be both obvious at the moment, as well as after the fact, if the car was being operated by its software or by the driver.
1: That's a possibly unintuitive offshoot of another principle in German law. Say you're a drunk pedestrian who stumbles into the street. A car swerves, hits a tree, and the car's passenger is killed. Who is at fault?
The somewhat unintuitive result is: the driver is at fault (meaning his insurance will have to pay for damages). The reasoning: Those travelling in cars derive all the value of car traffic. The pedestrian may have made a mistake–but they would never have been in a situation where it was possible to make that mistake, without others seeking the benefit of car traffic. Therefore, the pedestrian is never liable (unless, obviously, they act intentionally)
"The driver should be able to decide who owns his vehicle data."
This. This is the most important statement in the findings. We know accidents are going to happen, judgements made. The rest makes obvious, human centric, sense. (As a human that is. ;) )
8 comments
[ 3.6 ms ] story [ 22.7 ms ] threadTrue, but sounds like autonomous vehicles were a bitter pill to swallow for less accidents.
> In the case of unavoidable accidents, classification of drivers is prohibited. This includes determining fault by age, sex, or physical and mental constitution.
Does that the imply that we should only be in favour of auotonomous vehicles if they are always better than the worst drivers?
EDIT: I probably misunderstood this one. After reading the passage in the origal report I think it means that the computer is not allowed (for example) to prefer to save a child over an adult.
> We must be able to determine the responsible party in an accident at any time: human or computer.
Desirable, but good luck with that one.
> Drivers must have proper documentation.
Not sure what that means.
> The driver should be able to decide who owns his vehicle data.
Why not simply demand that the driver owns the data. If we leave the decision to the driver we will know how it ends up. They will assign away all their rights for a free cookie.
Most importantly the article (didn't read the report) skips one of the most important and difficult ethical questions. How the balance the life of other with the physical integrity of those in the car, especially in the light of unavoidable false positives.
Original Report (German): https://www.bmvi.de/SharedDocs/DE/Anlage/Presse/084-dobrindt...
The translation of the article isn't great. With "owning" data they mean to determine who has access to the data. The driver has to be able to deny companies using their data. If this is implemented it would be a good step, Tesla currently doesn't allow that.
Some of the less obvious thoughts:
- If it can be shown that self-driving systems will lead to fewer accidents, the government will have an obligation to allow them–even if some accidents may continue to happen.
- The focus should be on avoiding those situations where technology has to make morally ambiguous decisions
- Where such life-vs-life decisions do happen, it would be immoral to assign different values to the individual lives on the basis of sex, age or other characteristics.
- In principle, it would be possible to allow decisions trying to minimise the lives lost/put at risk, i. e. sacrificing one life to save two.
- But: they also deem it immoral to sacrifices the life of people who did not participate in the creation of the traffic risk themselves.(1) This means, for example, that system may not "sacrifice" one pedestrian for two (or even one) passengers of the car.
- Liability will move from the owner/operator to the producers of cars
- The technology, and the decisions it implements, must be communicated to the public
- It's unclear if traffic should be directed by some central authority. Such centralisation would need to find a way to operate in a way that makes it impossible to create profiles of individuals' movements
- The car's owner must always agree to the sharing and usage of their data. This requires not just the theoretically choice, but the real possibility to say no, and realistic options to do so without undue consequences.
- It must be both obvious at the moment, as well as after the fact, if the car was being operated by its software or by the driver.
1: That's a possibly unintuitive offshoot of another principle in German law. Say you're a drunk pedestrian who stumbles into the street. A car swerves, hits a tree, and the car's passenger is killed. Who is at fault?
The somewhat unintuitive result is: the driver is at fault (meaning his insurance will have to pay for damages). The reasoning: Those travelling in cars derive all the value of car traffic. The pedestrian may have made a mistake–but they would never have been in a situation where it was possible to make that mistake, without others seeking the benefit of car traffic. Therefore, the pedestrian is never liable (unless, obviously, they act intentionally)
https://en.wikipedia.org/wiki/Three_Laws_of_Robotics
This. This is the most important statement in the findings. We know accidents are going to happen, judgements made. The rest makes obvious, human centric, sense. (As a human that is. ;) )