ExplainabilityTransparencyAccountabilitycs.AI
Contextual Importance and Utility in Python: New Functionality and Insights with the py-ciu Package
Kary Främling
August 19, 2024
arXiv.org
Key Takeaways
-
Model-agnostic explanations are now more accessible: The py-ciu package provides researchers and practitioners with an easy-to-use tool for explaining AI decisions across different types of models, removing technical barriers that previously limited who could implement explainability methods in practice.
-
CIU goes beyond standard feature importance: The paper demonstrates that Contextual Importance and Utility offers capabilities beyond current state-of-the-art XAI methods, suggesting it can provide more nuanced insights into how AI models make decisions in specific contexts rather than just ranking which features matter most overall.
-
Post-hoc explanation methods enable broader adoption: By offering a post-hoc approach (analyzing models after they're built), CIU allows organizations to add explainability to existing AI systems without requiring model redesign, making responsible AI practices more feasible for real-world deployment.
-
Open-source implementation supports responsible AI development: The availability of reliable, easy-to-use software implementations is positioned as essential for both academic research and industry adoption of XAI methods, suggesting that accessibility of tools directly impacts how responsibly AI systems are developed and deployed.
Abstract
The availability of easy-to-use and reliable software implementations is important for allowing researchers in academia and industry to test, assess and take into use eXplainable AI (XAI) methods. This paper describes the \texttt{py-ciu} Python implementation of the Contextual Importance and Utility (CIU) model-agnostic, post-hoc explanation method and illustrates capabilities of CIU that go beyond the current state-of-the-art that could be useful for XAI practitioners in general.