Coarse ethics: how to ethically assess explainable artificial intelligence

Abstract The integration of artificial intelligence (AI) into human society mandates that their decision-making process is explicable to users, as exemplified in Asimov’s Three Laws Robotics. Such interpretability calls for explainable AI (XAI), which this paper cites various models. However, the transaction between computable accuracy and can be a trade-off, requiring answers questions about negotiable conditions degrees prediction may sacrificed enable user-interpretability. extant research has focussed on technical issues, but it also desirable apply branch ethics deal with trade-off problem. This scholarly domain labelled coarse study, discusses two issues vis-à-vis type evaluation. First, formal would allow trade-offs? study posits minimal requisites: adequately high coverage order-preservation. second issue concerns could justify interpretability, suggests justification methods: impracticability adjustment perspective from machine-computable human-interpretable. contributes by connecting autonomous systems future regulation formally assessing adequacy rationales.