Towards Kant Machines

Abstract

For some years now, ethics no longer only means human ethics. The young discipline of machine ethics researches the morality of semi-autonomous and autonomous systems like self-driving cars, robots and drones. Interactive software systems such as chatbots are also relevant. In 2013, the School of Business at the University of Applied Sciences and Arts Northwestern Switzerland FHNW implemented a prototype of the GOODBOT, which is a novelty chatbot and a simple moral machine. One of its meta-rules was that it should not lie unless not lying would hurt the user. In a follow-up project in 2016, the LIEBOT was developed, a kind of Munchausen machine. This article describes the back-ground and the foundations of this project and lists the chatbot’s strategies of lying. Then it discusses how Munchausen machines as immoral machines can contribute to the construction and optimization of moral machines, for example Kant machines, which prefer the truth. The LIEBOT serves as a contribution to machine ethics as well as a critical review of electronic language-based systems and services.