An ontology of explanation for social eXplainable Artificial Intelligence

The recent surge of interest in eXplainable Artificial Intelligence (XAI) has led to a growing focus on incorporating social insights into XAI and grounding explainability in social interaction. This has given rise to the concept of Social XAI. However, ontological modelling of the Social XAI framework remains limited. In particular, there is a lack of models that capture the progressivity of explanation processes in terms of the changes that occur. This paper introduces the Social XAI Ontology, a BFO-based reference ontology. The ontology models a variety of entities involved in the process of explanation. First, it provides a model of the participants in explanatory dialogues, such as artificial explainers and human explainees, and the roles they may take. Then, it introduces a rich set of explanatory moves that explainers and explainees perform. The paper further presents ontological accounts of four types of context that are central to real-world Social XAI projects. Furthermore, it provides an ontological account of different dimensions of adaptivity, which is considered to be an essential prerequisite for realizing Social XAI systems. Finally, the paper revisits two existing models of understanding in Social XAI by applying a recent theory of change. The paper concludes by discussing use cases for the Social XAI Ontology.