Explainability of Artificial Intelligence Models: Technical Foundations and Legal Principles

Abstract The now prevalent use of Artificial Intelligence (AI) and specifically machine learning driven models to automate the making decisions raises novel legal issues. One issue particular importance arises when rationale for automated decision is not readily determinable or traceable by virtue complexity model used: How can such a be legally assessed substantiated? any potential liability “wrong” properly determined? These questions are being explored organizations governments around world. A key informant analysis in these cases extent which question “explainable”. This paper seeks provide (1) an introductory overview technical components manner consumable someone without computer science mathematics background, (2) summary Canadian Vietnamese response explainability challenge so far, (3) what ”explanation” scientific domains, (4) preliminary framework analyzing sufficiency explanation its prediction(s).