Explainability of Complex AI Models with Correlation Impact Ratio

TL;DR

This paper introduces ExCIR, a new metric for explaining AI model predictions that effectively handles correlated features, is computationally efficient, and improves interpretability across diverse datasets.

Contribution

ExCIR provides a theoretically grounded, stable, and scalable method for feature importance explanation that overcomes limitations of existing post hoc explainers.

Findings

ExCIR outperforms existing methods in interpretability and stability.

It remains reliable under noise and sampling variations.

Validated across diverse datasets including EEG and image data.

Abstract

Complex AI systems make better predictions but often lack transparency, limiting trustworthiness, interpretability, and safe deployment. Common post hoc AI explainers, such as LIME, SHAP, HSIC, and SAGE, are model agnostic but are too restricted in one significant regard: they tend to misrank correlated features and require costly perturbations, which do not scale to high dimensional data. We introduce ExCIR (Explainability through Correlation Impact Ratio), a theoretically grounded, simple, and reliable metric for explaining the contribution of input features to model outputs, which remains stable and consistent under noise and sampling variations. We demonstrate that ExCIR captures dependencies arising from correlated features through a lightweight single pass formulation. Experimental evaluations on diverse datasets, including EEG, synthetic vehicular data, Digits, and Cats-Dogs, validate the effectiveness and stability of ExCIR across domains, achieving more interpretable feature explanations than existing methods while remaining computationally efficient. To this end, we further extend ExCIR with an information theoretic foundation that unifies the correlation ratio with Canonical Correlation Analysis under mutual information bounds, enabling multi output and class conditioned explainability at scale.