Towards an Ontology of Explanations
Abstract
The ability of an agent to explain how it arrived at a decision is important for garnering trust in the agent, understanding its operation, and possibly gaining new and more generalizable knowledge. With AI agents becoming more complex and widespread, there is growing interest in providing them with the ability to explain themselves. Explanations, however, are more than a matter of integrating techniques to approximate machine learning models with more straightforward techniques; explanations are interactive communication acts which must be tailored to the interests of the person seeking explanations. To better represent the communication goals behind an explanation, we propose the “Explanation Interchange Format” ontology, which has several aims. First, it formally describes the communication act of Explanation and its structure based on existing theoretical work on scientific explanations. Second, the ontology enables reasoning to construct explanations, e.g. selecting explanatory structures appropriate to a questioner’s interests. We illustrate reasoning on our ontology with some examples.
Full citation
"Towards an Ontology of Explanations", Groza, Adrian and Pomarlan, Mihai, Measuring Ontologies for Value Enhancement: Aligning Computing Productivity with Human Creativity for Societal Adaptation Springer Nature Switzerland, pp. 73--85, Cham, DOI: https://doi.org/10.1007/978-3-031-22228-3_4, ISBN: 978-3-031-22228-3, 2023.