The Need for Ethical Artificial Intelligence and the Question of Embedding Moral and Ethical Principles

Abstract

The issue of Facebook moderators made topical headlines in 2019. When confronted with horrifying images such as terrorist assaults filmed online, moderators are required to be extremely reactive in order to eliminate the scenes violating human dignity from the social network as rapidly as possible. In [1], my colleagues and I explored whether it was possible or not to model human values such as virtue.Modeling could eventually make it possible to automatize all or part of the moderators’ arduous and ungrateful work. In a first part, this article deals with the need for a reflection on ethical Artificial Intelligence (AI). After providing a definition of what AI is, I then discuss how ethical rules could be implemented in a system using AI. In a second part, I ask myself whether it is possible to embed moral and ethical principles. Using a utility function can help agents to make ethical decisions, but it is also important to be cautious about the limitations of such a sometimes-simplistic approach. Ethics must be appreciated in the social and technical context in which it is first developed and then implemented [2].