AI as Moral Device – A reflection on roles and distance

Ongoing research efforts in the field of engineering and HCI have investigated ways of operationalising human moral decision-making processes and replicate these mechanics in algorithms [3] with the aim of creating independent artificial moral agents. While persuasive and nudging technologies have been successful in creating environments that support certain behaviours while limiting or inhibiting others, second-order ethical design, the design of systems that take and enact ethical stances, is facing challenges on many fronts, posing the question whether true moral artificial agents are indeed possible. While algorithmic machine learning systems are already in use in morally relevant fields, such as in criminal justice cases and to allocate financial opportunities, the question remains whether these agents are considered as possessing their own morality, simply because professionals base their moral judgements on their recommendations. In this position paper I want to discuss whether the creation of artificial moral agents is actually desirable and argue that current efforts to that goal might lead in fact to an increased distancing of people from moral deliberation processes, in particular when it comes to data-driven algorithmic systems, which are considered to embody AI