Community-in-the-loop

Today, due to growing computing power and the increasing availability of high-quality datasets, artificial intelligence (AI) technologies are entering many areas of our everyday life. Thereby, however, significant ethical concerns arise, including issues of fairness, privacy and human autonomy. By aggregating current concerns and criticisms, we identify five crucial shortcomings of the current debate on the ethics of AI. On the threshold of a third wave of AI ethics, we find that the field eventually fails to take sufficient account of the business context and deep societal value conflicts the use of AI systems may evoke. For even a perfectly fair AI system, regardless of its feasibility, may be ethically problematic, a too narrow focus on the ethical implications of technical systems alone seems insufficient. Therefore, we introduce a business ethics perspective based on the normative theory of contractualism and conceptualise ethical implications as conflicts between values of diverse stakeholders. We argue that such value conflicts can be resolved by an account of deliberative order ethics holding that stakeholders of an economic community deliberate the costs and benefits and agree on rules for acceptable trade-offs when AI systems are employed. This allows AI ethics to consider business practices, to recognise the role of firms, and ethical AI not being at risk to provide a competitive disadvantage or in conflict with the current functioning of economic markets. By introducing deliberative order ethics, we thus seek to do justice to the fundamental normative and political dimensions at the core of AI ethics.