Explanation as a Process: User-Centric Construction of Multi-level and Multi-modal Explanations

In the last years, XAI research has mainly been concerned with developing new technical approaches to explain deep learning models. Just recent started acknowledge need tailor explanations different contexts and requirements of stakeholders. Explanations must not only suit developers models, but also domain experts as well end users. Thus, in order satisfy stakeholders, explanation methods be combined. While multi-modal have used make model predictions more transparent, less focused on treating a process, where users can ask for information according level understanding gained at certain point time. Consequently, an opportunity explore levels abstraction should provided besides explanations. We present process-based approach that combines multi-level The user textual or visualizations through conversational interaction drill-down manner. use Inductive Logic Programming, interpretable machine approach, learn comprehensible model. Further, we algorithm creates explanatory tree each example which classifier decision is explained. navigated by get answers detail. provide proof-of-concept implementation concepts induced from semantic net about living beings.