First published July 1 2016
A Bottom-Up Approach to Machine Ethics
This paper presents a bottom-up approach to machine ethics, based on the Measurement Logic Machine (MLM). It is explained how ethical notions emerge from the workings, architecture, and environmental assumptions of the MLM framework. The MLM uses sequences of measurements to perform short-term predictive inference. The MLM ethical behavior stems from the inner evaluation of measurements that are used to filter the predictions. The MLM ethical discernment is based on measurements that detect immediate suffering in other agents. Also, a definition of what is an ethically positive modification of the inner evaluations is proposed, based on the notion of environmental intelligence and the corresponding notion of suffering. It is shown how this double approach is consistent with our intuitive notion of ethics. The MLM, with or without ethical discernment, can be used in evolutionary game theory, and gives clues to the search of ethical senses that increase the chances of survival of autonomous agents.