Towards a Relational Ethics in AI. The Problem of Agency, The Search for Common Principles, the Pairing of Human and Artificial Agents

In this paper, some of the core problems of AI ethics are discussed. Firstly, assuming that AI ethics is to be an ethics of relations between human and artificial agents, we try to explain what “acting” means with respect to humans and AI systems. To do so, we outline the different forms of relationships that can be built and regulated between human and artificial agents, also according to a deontological approach. Secondly, then, we give an insight into the regulatory codes and documents produced across the world. We discuss their scope of applicability and global acceptance. In particular, we note that, along with principles of transparency, responsibility, privacy etc., some relational ethical principle is missing, although seeming essential. Thirdly, as a consequence, trying to concretely build upon such relational principle, we lay down the concept of a new kind of artificial intelligence systems design called endless tuning, which can help to develop and procedurally implement a common, cooperative and responsible conduct for humans and artificial entities alike.