No Explanation Without (Fuzzy) Representation

Explainability is seen as a necessary step in the adoption of AI, and is a building block for collaborative AI systems which combine the strengths of machines and humans in tackling problems such as the rapid analysis of cyber-security data. In this paper we argue that symbolic representations are a necessary component in the interface between humans and AI components for both explanation and the wider goal of collaborative systems. We show that fuzzy conceptual graphs are a feasible representation of general and specific knowledge in the domain of cyber-security, and illustrate that reasoning can enable the automatic generation of new knowledge.