Formalising Explainability and Interpretability: A Conceptual Framework and Taxonomical Dimensions
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Research on explainable artificial intelligence (XAI) has expanded rapidly, yet the concepts of interpretability and explainability remain theoretically unstable and inconsistently defined. This paper addresses that weakness directly through a systematic literature review focused specifically on published definitions of the two concepts. We analyse these definitions by decomposing them into semantic units and reorganising the recurring conceptual material into a structured framework centred on four entities: AI system, Human user, Interaction, and Context. On this basis, we derive a set of high-level taxonomical dimensions along which existing accounts differ, including their formalism, measurability, specificity, and dependence on users, domains, and contexts. Our analysis shows that interpretability and explainability are not cleanly separable in the literature, but are better understood as overlapping regions of a broader multidimensional conceptual space. The paper therefore contributes a methodological approach for semantic analysis in fragmented research areas, a novel conceptual framework for interpretability and explainability, and a taxonomical basis for comparing existing definitions. More broadly, it takes a concrete step towards the formalisation of core XAI concepts and towards a more mature theoretical foundation for the field.
Fields of Study & Subjects
Explainable Artificial Intelligence (XAI)
Ethics and Social Impacts of AI
Safety Systems Engineering in Autonomy
Publication details
- Journal / source
- Research Square
- Publication date
- 31 March 2026
- Language
- EN
- Document type
- Preprint