Due to ongoing technological advancement, self-driving cars already navigate our roads. This raises the urgent question of the ethically correct programming of artificial intelligence.
While in Austria the topic of ethical programming of self-driving cars is hardly dealt with, in Germany an expert team has been set up for ethical issues related to self-driving cars. Under the direction of the former German Federal Constitutional Court Judge Udo di Fabio various guidelines for the programming of automated driving systems were developed.
The proper programming of self-driving cars and the interconnected artificial intelligence is fundamental when it comes to the decision-making capabilities said car needs in critical driving situations. This requires the development of an evaluating system, especially if an accident cannot be avoided. In the event of an emerging probability of accident the artificial intelligence needs to decide, who or what may be affected by an accident. According to a report from expert programmers of automated driving systems, the basic principles are relatively simple: it seems understandable that in cases of a heightened risk of accident the decision-making-process must be programmed to prioritize property damage over personal injury. In a dangerous situation, the computer-controlled self-driving car has to give human life the highest priority. To avoid the imminent danger of a person being run over and thus potentially being injured or killed, the computer must initiate an evasive maneuver, which will most likely only lead to property damage.
The question of choosing between lives poses a greater challenge: for example, when a self-driving car has to choose between two accident situations in which people will be injured or die either way. The German Ethics Committee has developed a principle which states that every human life must be equally worth saving; it is thereby not permissible to program identifying-systems that assess the value of a human life. Neither the physical or mental state, nor age or sex may be indicators for an assessment or selection of behavior during an accident situation. However, this only applies to the singular identification of the individual victim, but not to the quantity of casualties. Thus, the ethics committee has decided that it is entirely eligible to establish an assessment according to the number of potential accident victims. It is acceptable to jeopardize a single accident victim in order to protect two or more potentially at-risk people. The number of potential victims is certainly an aspect that needs to be taken into consideration.
Differentiating between the people involved in the accident seems to be an even greater problem: the automobile industry’s biggest interest is of course to protect their customers, the drivers and passengers of their vehicles. This raises the question of whether it is permissible for a manufacturer to program a prioritization in saving the life of the driver rather than that of a driver in another car or even a pedestrian. While at first glance such programming may seem to be inconsistent with our ethical standards, this case becomes even more complicated when the question of correct behavior in traffic is raised.
If, for example, a pedestrian suddenly decides to illegally cross a red light, should the driver be obligated to risk an evasive maneuver and endanger himself? Without the support of artificial intelligence, this issue is incumbent upon the driver's reflexive behavior and possibly a subsequent assessment by the courts. During the process of programming artificial intelligence for self-driving cars accordingly, however, such evaluations must be taken into account. Therefore, the German panel of experts argues that it is permissible to prioritize traffic participants who have proven to behave correctly over those who do not comply with the traffic rules.
Even if the assessment system of accident situations remains a complex ethical and legal issue, it is believed that society as a whole can benefit from self-driving cars. Studies assume that most car accidents are caused by human error and that many accidents could be avoided if we rely on artificial intelligence and make automated driving systems our first choice when entering a car.