Key Takeaways on Ontologies and Explainable AI

• Ontologies provide structured knowledge frameworks that can make AI systems more transparent by organizing information in ways humans can understand, addressing a core challenge in responsible AI: the need for intelligible explanations rather than "black box" outputs.

• Three practical applications exist for ontologies in AI: reference modelling (creating shared understanding of concepts), common-sense reasoning (enabling AI to apply everyday logic), and knowledge refinement (managing system complexity)—each contributing differently to making AI decisions more explainable.

• Human-understandability remains an open challenge: The paper identifies that while ontology-based approaches show promise for explainability, significant work is still needed to evaluate whether these explanations actually help humans understand AI decisions effectively, highlighting a gap between technical explainability and real-world usability.

• Ontologies support human-centric AI development by enabling systems to reason about knowledge in structured ways that align with how humans conceptualize problems, potentially bridging the gap between AI logic and human intuition in responsible AI systems.

• Evaluation of explanation effectiveness is underdeveloped: The paper concludes that the field lacks robust methods to measure whether ontology-based explanations genuinely improve human comprehension and trust in AI systems, suggesting this is a critical area for future responsible AI research.