When trusted black boxes don't agree: Incentivizing iterative improvement and accountability in critical software systems

Publication
Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society