Traceability and accountability in autonomous agents

Abstract

The ability to understand why a particular robot behavior was triggered is a cornerstone for having human-acceptable social robots. Every robot action should be able to be explained and audited, expected and unexpected robot behaviors should generate a fingerprint showing the components and events that produce them. This research proposes a three-dimensional model for accountability in autonomous robots. The model shows three different elements that deal with the different levels of information, from low-level such as logging to a high level, such as robot behavior. The model proposes three different approaches of visualization, one on command line and two using GUI. The three-level system allows a better understanding of robot behaviors and simplifies the mapping of observable behaviors with accountable information.