Explainable Artificial Intelligence and Multicollinearity : A Mini Review of Current Approaches
Ahmed M Salih
TL;DR
Multicollinearity undermines the reliability of XAI explanations, particularly for feature-importance and dependence plots. This mini-review synthesizes state-of-the-art approaches to address multicollinearity in XAI, focusing on seven methods (MIP, Extended Kernel SHAP, NMR, SCR, MCC, CS, CIT) and their global/local applicability. It highlights that most existing solutions are SHAP-centric or local, with limitations tied to sampling efficiency and high-dimensionality, and notes the absence of a general, intrinsic mitigation across XAI methods. The discussion emphasizes the need for probabilistic and visualization frameworks that reflect interactions among correlated features, to improve XAI robustness in domains with correlated data such as healthcare and biology. By outlining current gaps and future directions, the review guides researchers toward more reliable explanations in multicollinear settings.
Abstract
Explainable Artificial Intelligence (XAI) methods help to understand the internal mechanism of machine learning models and how they reach a specific decision or made a specific action. The list of informative features is one of the most common output of XAI methods. Multicollinearity is one of the big issue that should be considered when XAI generates the explanation in terms of the most informative features in an AI system. No review has been dedicated to investigate the current approaches to handle such significant issue. In this paper, we provide a review of the current state-of-the-art approaches in relation to the XAI in the context of recent advances in dealing with the multicollinearity issue. To do so, we searched in three repositories that are: Web of Science, Scopus and IEEE Xplore to find pertinent published papers. After excluding irrelevant papers, seven papers were considered in the review. In addition, we discuss the current XAI methods and their limitations in dealing with the multicollinearity and suggest future directions.