Approaches to Forming ETHICAL AI as an Artificial Moral Agent: Suggesting Virtue Education Method Through Comparison of Top-Down and Bottom-Up Approaches

Abstract

Purpose: The purpose of this study is to propose that the virtue education method can be utilized in the process of learning to have ethics as an Artificial Moral Agent(AMA). Recent various discussions about AI presuppose the coexistence of humans and AI. Meanwhile, in order for coexistence to be achieved, it is fundamentally necessary for humans as individual beings to form relationships with others in an appropriate way. This study focused on the use of the virtue education method among the methods for AI to learn the order of relationships as an AMA that becomes the counterpart of human ethical behavior in the interaction between humans and AI. Method: This study presented basic data using the method of literature research. And based on this, a virtue education method that can be applied to AMA was developed and presented using the method of development research. With this, This research tried to argue about the necessity and validity of the virtue education method. Results: This study examines that the current discussion on ethical learning in AMA is mainly based on a Top-down approach based on traditional ethical theory. And as an alternative to this, it is suggested that machine learning and deep learning methods can be used as a method for AMA to learn ethics. Based on this, the characteristics and validity of virtue ethics applicable to AMA were presented. And as a method to learn it, a teaching and learning method applicable to the learning of AMA was proposed. Conclusion: Currently, the approach to preparing ethics for AMA tends to mainly choose a Top-down approach in which a specific ethical theory is selected and the design of a computational algorithm and system that can implement the theory is derived. However, just as children develop moral reasoning through appropriate developmental stages, AMA also needs to actively introduce a bottom-up approach to forming ethical judgment principles. As a specific method, we propose to use the method of grafting the virtue education method of analyzing the characteristics of a virtuous person and repeating it by imitating it to the deep learning method of AI.