The introduction of networked AI-robots into society is estimated to bring numerous ethical issues to the fore. Moreover, there is a possibility that conventional ethics, such as engineering ethics and information ethics may not sufficiently address these issues. Hence, a new field called machine ethics is emerging that introduces ethical judgment and behavioral mechanisms into AI and robots. Accordingly, various technical elements are being studied to empower machines to make ethical decisions. Previously, symbolic reasoning systems were the main focus of this field; however, in recent years neural network models have also been utilized. In this paper, the limitations of conventional applied ethics are considered from the ethical viewpoint of the relationship between the autonomy and network complexity of networked AI-robots. Therefore, the possibility of multi-agent type deep reinforcement learning is considered as a method for overcoming the encountered limitations. From this, a technical frame for emergence of ethical norms in the networked AI-robot society is proposed.
Machine Ethics
Ethical norms
Deep Reinforcement Learning
Multi-agent system