Explanation, semantics, and ontology

Abstract

The terms ‘semantics’ and ‘ontology’ are increasingly appearing together with ‘explanation’, not only in the scientific literature, but also in everyday social interactions, in particular, within organizations. Ontologies have been shown to play a key role in supporting the semantic interoperability of data and knowledge representation structures used by information systems. With the proliferation of applications of Artificial Intelligence (AI) in different settings and the increasing need to guarantee their explainability (but also their interoperability) in critical contexts, the term ‘explanation’ has also become part of the scientific and technical jargon of modern information systems engineering. However, all of these terms are also significantly overloaded. In this paper, we address several interpretations of these notions, with an emphasis on their strong connection. Specifically, we discuss a notion of explanation termed ontological unpacking, which aims at explaining symbolic domain descriptions (e.g., conceptual models, knowledge graphs, logical specifications) by revealing their ontological commitment in terms of their so-called truthmakers, i.e., the entities in one’s ontology that are responsible for the truth of a description. To illustrate this methodology, we employ an ontological theory of relations to explain a symbolic model encoded in the de facto standard modeling language UML. We also discuss the essential role played by ontology-driven conceptual models (resulting from this form of explanation processes) in supporting semantic interoperability tasks. Furthermore, we revisit a proposal for quality criteria for explanations from philosophy of science to assess our approach. Finally, we discuss the relation between ontological unpacking and other forms of explanation in philosophy and science, as well as in the subarea of Artificial Intelligence known as Explainable AI (XAI).