ABSTRACT

This chapter discusses the risk factors of robotic moral agents. The main factors that will lead robots to moral failure are 1) failure to ground symbols correctly, 2) inadequate or incorrect rules for specific norm tests and 3) inadequate or incorrect weighting of rules in reasonable person tests. The risks of moral failure in robots can be mitigated by functional, regression and other forms of testing. Mesh architecture is introduced as a way to mitigate the risks of robotic moral failure. The essential feature of mesh architecture is the elimination of single points of failure.