Ethical Markov Decision Processes with Moral Worth as Rewards

Mihail Stojanovski (Normandie Univ, UNICAEN, ENSICAEN, CNRS, GREYC), Nadjet Bourdache (Normandie Univ, UNICAEN, ENSICAEN, CNRS, GREYC), Grégory Bonnet (Normandie Univ, UNICAEN, ENSICAEN, CNRS, GREYC), Abdel-Illah Mouaddib (Normandie Univ, UNICAEN, ENSICAEN, CNRS, GREYC)

Abstract

We propose an expressive framework for specifying ethical behaviours, called Ethical Markov Decision Processes (E-MDPs) that extends classical MDPs with the explicit representation of moral values-positive or negative-that the agent's decisions may promote or demote.