We analyze opportunities and challenges for explainable artificial intelligence in molecular design, emphasize chemically intuitive explanations of predictions, and discuss advantages of including domain-specific knowledge.
Lamens et al. (Thu,) studied this question.