Explanations Based on Item Response Theory (eXirt): A Model-Specific Method to Explain Tree-Ensemble Model in Trust Perspective

TL;DR

This work introduces eXirt, a model-specific XAI method that leverages Item Response Theory to explain tree-ensemble models on tabular data. It yields global feature relevance ranks and local explanations via Item Characteristic Curves and Explanation-by-Example, enabling insight into model reliability and trust. The approach is validated across 41 binary-class datasets and four algorithms, compared against six established XAI methods, and analyzed through dataset clustering and Spearman rank correlations. Results show that eXirt often provides explanations that diverge from existing methods, highlighting unique perspectives on model behavior and offering a robust basis for trust in predictions. The study also demonstrates how local ICC-based explanations can reveal instance-level reliability and potential biases, pointing to practical benefits for deploying high-performing yet interpretable tree-ensembles.

Abstract

In recent years, XAI researchers have been formalizing proposals and developing new methods to explain black box models, with no general consensus in the community on which method to use to explain these models, with this choice being almost directly linked to the popularity of a specific method. Methods such as Ciu, Dalex, Eli5, Lofo, Shap and Skater emerged with the proposal to explain black box models through global rankings of feature relevance, which based on different methodologies, generate global explanations that indicate how the model's inputs explain its predictions. In this context, 41 datasets, 4 tree-ensemble algorithms (Light Gradient Boosting, CatBoost, Random Forest, and Gradient Boosting), and 6 XAI methods were used to support the launch of a new XAI method, called eXirt, based on Item Response Theory - IRT and aimed at tree-ensemble black box models that use tabular data referring to binary classification problems. In the first set of analyses, the 164 global feature relevance ranks of the eXirt were compared with 984 ranks of the other XAI methods present in the literature, seeking to highlight their similarities and differences. In a second analysis, exclusive explanations of the eXirt based on Explanation-by-example were presented that help in understanding the model trust. Thus, it was verified that eXirt is able to generate global explanations of tree-ensemble models and also local explanations of instances of models through IRT, showing how this consolidated theory can be used in machine learning in order to obtain explainable and reliable models.

Figures (16)

• Figure 1: Example of the representation of the parameter values of an item arranged on the Item Characteristic Curve - ICC. The letters $a$, $b$ and $c$ represent the discrimination, difficulty and guessing properties, respectively.
• Figure 2: Associations of IRT terms and Machine Learning terms.
• Figure 3: Visual scheme of all steps and processes performed by the pipeline.
• Figure 4: Silhouette coefficients for clustering, using the K-means algorithm, for $K = 4$. Distance means (axis $x$) and label of clusters $0, 1, 2$ and $3$ (axis $y$).
• Figure 5: Visual summary of all steps and processes performed by the XAI measure, called eXirt.