We argue in this paper that the idea of a "moral Turing test" is insufficient for an ethical evaluation of an autonomous system's behavior. We propose that verification, where the behavior of a system is designed for transparency and explicitly represents decision-making based on norms, is a better standard of evaluation.
@article{arnold2016against,
title={Against the moral Turing test: accountable design and the moral reasoning of autonomous systems},
author={Thomas Arnold and Matthias Scheutz},
year={2016},
journal={Ethics and Information Technology},
publisher={Springer},
volume={18},
pages={103--115}
url={https://hrilab.tufts.edu/publications/arnold2016against.pdf}
}