ExplainabilityTransparencyAccountabilitycs.LG
The Need for Standardized Explainability
Othman Benchekroun, Adel Rahimi, Qini Zhang, Tetiana Kodliuk
October 20, 2020
arXiv.org
11 citations
2 influential
Key Takeaways
-
Explainability is critical but currently inconsistent: The paper identifies that while explainable AI (XAI) is essential for real-world AI systems, existing methods lack standardization, making it difficult to ensure AI decisions can be reliably understood across different applications and organizations.
-
Clear definitions are needed to move forward: The authors propose novel standardized definitions for "Explainability" and "Interpretability" as a foundational step toward creating consistency in how the field approaches making AI systems transparent and understandable.
-
The field needs a common framework: By offering a tentative taxonomy of different explainability methods, the paper suggests that organizing existing approaches into a structured classification system could help practitioners choose appropriate methods and enable better comparison across different AI systems.
-
Current methods are insufficient for industry needs: The paper emphasizes that existing explainability techniques fail to adequately address the practical requirements of industry-grade AI systems, suggesting that standardization could help bridge this gap between research and real-world deployment.
Abstract
Explainable AI (XAI) is paramount in industry-grade AI; however existing methods fail to address this necessity, in part due to a lack of standardisation of explainability methods. The purpose of this paper is to offer a perspective on the current state of the area of explainability, and to provide novel definitions for Explainability and Interpretability to begin standardising this area of research. To do so, we provide an overview of the literature on explainability, and of the existing methods that are already implemented. Finally, we offer a tentative taxonomy of the different explainability methods, opening the door to future research.