Towards Human-Compatible XAI: Explaining Data Differentials with Concept Induction over Background Knowledge