Knowledge Graphs and XAI Enhance Urban Mining Decision Defensibility

This research explores how the integration of Knowledge Graphs (KGs) and Explainable AI (XAI) can significantly enhance the pre-demolition assessment process in urban mining. The core argument is that for AI support in this regulated field, the primary value lies not just in predictive accuracy, but in the defensibility of decisions—meaning they must be legible, plausible, traceable to sources, and contestable. While KGs provide structured domain knowledge and XAI offers insights into model reasoning, neither fully addresses these requirements independently. The paper introduces a complementarity-theoretic interpretation, defining four specific KG-XAI integration modes: Lifting, Constraining, Typing, and Revising. Each mode is characterized as a distinct operation that combines XAI artifacts with KG structures, unlocking unique properties crucial for decision defensibility. An example from urban mining, using a fire-door scenario and W3C Linked Building Data, illustrates how these integrated modes produce regulatory artifacts that are more robust and accountable than those generated by either technology alone.