LaPLACE: Probabilistic Local Model-Agnostic Causal Explanations

TL;DR

LaPLACE is a probabilistic, model-agnostic explanation method for classifiers on tabular data that uses Markov blankets to generate flexible, causal explanations without fixed feature limits, outperforming existing explainers.

Contribution

Introduces LaPLACE-explainer, a novel probabilistic causal explanation method leveraging Markov blankets, eliminating fixed feature count constraints and improving accuracy and consistency over LIME and SHAP.

Findings

Outperforms LIME and SHAP in local accuracy and feature consistency

Provides probabilistic causal explanations with automatic feature subset selection

Validated on simulated and real-world datasets for trust and fairness insights

Abstract

Machine learning models have undeniably achieved impressive performance across a range of applications. However, their often perceived black-box nature, and lack of transparency in decision-making, have raised concerns about understanding their predictions. To tackle this challenge, researchers have developed methods to provide explanations for machine learning models. In this paper, we introduce LaPLACE-explainer, designed to provide probabilistic cause-and-effect explanations for any classifier operating on tabular data, in a human-understandable manner. The LaPLACE-Explainer component leverages the concept of a Markov blanket to establish statistical boundaries between relevant and non-relevant features automatically. This approach results in the automatic generation of optimal feature subsets, serving as explanations for predictions. Importantly, this eliminates the need to predetermine a fixed number N of top features as explanations, enhancing the flexibility and adaptability of our methodology. Through the incorporation of conditional probabilities, our approach offers probabilistic causal explanations and outperforms LIME and SHAP (well-known model-agnostic explainers) in terms of local accuracy and consistency of explained features. LaPLACE's soundness, consistency, local accuracy, and adaptability are rigorously validated across various classification models. Furthermore, we demonstrate the practical utility of these explanations via experiments with both simulated and real-world datasets. This encompasses addressing trust-related issues, such as evaluating prediction reliability, facilitating model selection, enhancing trustworthiness, and identifying fairness-related concerns within classifiers.