Submitted: August 2018

Abstract

We find ourselves surrounded by a rapidly increasing number of autonomous and semi-autonomous systems. Two great challenges arise from this development: Machine Ethics and Machine Explainability. Machine Ethics is concerned with behavioral constraints for systems so that morally acceptable, restricted behavior emerges; Machine Explainability provides the possibility for the actions and decisions of systems to be explained satisfactorily, so that human users can understand and justifiably trust those systems.

In this master’s thesis, I want to emphasize the need to link and cross-fertilize between these two areas. I point out how Machine Ethics calls for Machine Explainability and how Machine Explainability involves Machine Ethics.

I develop both these facets based on a toy example from the context of medical care robots. In this context, I argue that moral behavior, even if it were verifiable and verified, is insufficient to establish justified trust in an autonomous system. The system must be supplemented with the ability for its decisions to be explained or even with the ability to explain its decisions itself. It should thus be supplemented by a Machine Explanation component. Conversely, generated explanations must refer to the system’s model- and constraint-based Machine Ethics reasoning.

I propose to use arguments for the task of generating useful explanations of the Machine Explanation component, and I sketch out how the content of the arguments must use the moral reasoning of the Machine Ethics component. Furthermore, I examine how it might be possible to explain systems without a Machine Explanation component.