Explainability Engineering Challenges: Connecting Explainability Levels to Run-Time Explainability

While automated and intelligent software systems are more and more used in everyday software systems, we must ensure that these systems remain understandable to all involved stakeholders. For this, two disciplines can benefit from each other: Explainability engineering, which integrates explainability into classical software systems engineering methodologies, and eXplainable Artificial Intelligence (XAI), which provides explanations for opaque AI system components. We discuss how to integrate levels of explainability requirements into an approach for run-time explainability as a core explainability engineering topic. For this, local and global explanations must be distinguished. We further on motivate the crucial role of XAI in explainability engineering and discuss challenges for bringing together the disciplines.