Improving AI systems' dependability by utilizing historical knowledge | IEEE Conference Publication | IEEE Xplore