Autonomous Cars

The use of cars automatically driven by software and not by a human being.

Examples

  • Robo-taxis – autonomous cars owned by a private company to transport people in urban environments
  • Autonomous cars controlled by a person from outside used for humanitarian missions, military operations or transport of persons or goods in extreme terrains
  • Examples and issues related to robots in transportation are discussed in (COMEST, 2017)

Benefits

  • Avoiding accidents originated by fatigue or distraction of a human driver 
  • Contributing to improving traffic order
  • Enabling independent mobility for persons with disabilities 

Threats related to misuse and abuse

Defining responsibility for accidents – Autonomous cars, as those driven by persons, can crash into another car, a pedestrian or a cyclist. Whose responsibility is it? Is the responsibility of the owner of the car, the software developer, the producer, all of them? 

Making decisions for unforeseen situations – It may happen that the car faces a number of unforeseen situations while driving autonomously; for example, another car crossing a street with red light, a pedestrian crossing the street when the traffic light allows cars to move, need for avoiding a street pothole, a car not keeping its line, the need for immediately stopping the car while circulating at speed and without enough distance to do it, and others. What can be considered the correct decisions to make? How will cars make the correct decisions? 

Hacking the software – An autonomous car is a software and hardware system that can be hacked. How do we prevent the car software for being hacked? How can responsibility for hacking car software be attributed? 

Violation of driver and passenger´s privacy – An autonomous car has sensors, hardware and software enabling the car to drive automatically. However, other functionality can be implemented without the users of the car being aware of this fact. Is the software of the car open to facilitate auditing of the implemented functionality? Is data about the car movements being registered? Is data about passengers of the car being registered? Is the behavior of passengers inside the car monitored? By whom?

Restricting areas of circulation – Given the use of geo-referential systems, the car could decide automatically the route to commute from one place to another. What will be the criteria to decide how to move from one place to another? Can a person force the algorithm to go or not to go through a certain area? If the decision is made automatically based on the shortest route, or safest route, or other criteria, is there a possibility for discriminating specific neighborhoods or locations? Is there a difference between personal decisions not to circulate through an area, compared with the decision made by an algorithm? Can the algorithm have a pre-determined bias to select some given routes? Who are benefiting and who are suffering from such pre-determined bias?

Diverse cultural aspects for automating decision-making – A study conducted by the Massachusetts Institute of Technology (MIT) collected responses among 40 million people in 233 countries on what an autonomous car should do when facing a moral dilemma. It revealed that sociocultural origin greatly influences opinion. From the analysis of the collected responses, researchers have discovered that people’s preferences differ widely depending on the country, but are highly correlated with culture and economy. The study defines interesting effects for countries that are currently conducting tests with autonomous cars, since these preferences could influence the configuration of the design and regulation of such vehicles. Car manufacturers should be aware, for example, that Chinese consumers would prefer a vehicle that protects the occupant rather than pedestrians. Thus, the design of the algorithm includes ethical and moral aspects. Do decisions made to resolve a moral dilemma (e.g. crashing into a pedestrian or crashing into another car) depend on the local culture? If so, how do car manufacturers implement different decision-making algorithms based on cultures?

Ethical challenges

The following table summarizes ethical challenges associated with the use of autonomous cars.

ID CHALLENGE RELATED TO *
C1 Deciding the person/entity responsible for accidents caused by autonomous cars Principle M Principle O
C2 Ensuring that the driving algorithm will make the correct decision when an autonomous car faces an unforeseen situation Principle O
C3 Avoiding hacking with malicious intentions autonomous car software Principle M
C4 Ensuring that the different electronic devices deployed in an autonomous car protect driver and passenger´s privacy – i.e. do not record unauthorized data about them Principle R Principle M
C5 Ensuring transparency and absence of bias when automatically deciding the routes that the autonomous car will follow Principle O Principle A
C6  Ensuring adherence to diverse cultural principles in automated driving algorithms Principle O

* See Principles for more information about principles