Is there a special form of ethics for artificial intelligence? Digital ethics for software products already exists – doesn’t it apply to AI? In principle, yes, but AI as a high-impact technology raises questions that have been more or less irrelevant so far.
Autonomous systems have become a much-discussed topic. The question probably raised most often concerns the decisions an autonomous vehicle may make in an accident. Let’s say a car approaches a crossing and a child runs in front of it. Should the car swerve to avoid hitting the child, if it then hits an adult waiting for the traffic lights to change? The MIT Moral Machine research project gathered human perspectives on such questions. Some of its results were published in the Nature magazine (Nature 563, 2018, p. 59-64). They show interesting cultural differences in attitudes towards young and old people in different countries.
The example mentioned above demonstrates the nature of ethical questions. In most cases, there is no right answer; instead, the problem needs to be addressed from various angles. In many fields of education, such as engineering, students are used to dealing with mathematical truths and learning facts rather than considering different opinions. Consequently, they may find ethical questions difficult or even boring. This is all the more reason to include ethics in studies in fields such as engineering.
The principles of law provide a starting point for ethical considerations. Human rights and equality issues are of particular importance, as demonstrated in the decision made by the National Non-Discrimination and Equality Tribunal of Finland in April 2018. It concluded that an automated consumer credit granting system had discriminated against the appellant because the decision was based only on background data such as age, gender, place of residence and native language.
The ethics of artificial intelligence have been discussed by many parties, such as the EU High-Level Expert Group on Artificial Intelligence. Typically the parties produce a list of things to remember, to be applied on a case-by-case basis. Such lists are recommendations rather than mandatory requirements.
Inspired by the enormous success of the Elements of AI course, the University of Helsinki is now preparing an online course for the general public on the ethics of AI. During this course, artificial intelligence will be discussed from the ethical, moral and legal perspectives, with case studies offering a more concrete approach. The launch of this online course is scheduled for autumn.