Formal Definition of Interpretability and Explainability in XAI

The interpretability and explicability of machine learning models lack a singular, formal definition, with understanding varying based on context and specific needs. This article conducts an in-depth review of current literature, aiming to provide a comprehensive synthesis of diverse perspectives and approaches within this dynamic field. A major contribution of this article is the proposal of a unified formal definition for interpretability and explicability. This proposal results from a careful synthesis of perspectives from the literature, integrating essential aspects such as human comprehensibility, justifiability, and clarity in presenting explanations. The fundamental goal is to consolidate knowledge on interpretability and explicability, offering a formal definition as a basis for a more coherent and unified understanding in machine learning. This unified definition is crucial for guiding future research efforts and advancing the field towards more transparent and understandable machine learning models.