CXPlain: Causal Explanations for Model Interpretation under Uncertainty

TL;DR

CXPlain introduces a causal learning approach to rapidly generate accurate feature importance explanations for machine-learning models while quantifying uncertainty, outperforming existing methods.

Contribution

The paper presents CXPlain, a novel causal explanation model that provides fast, accurate feature importance estimates and uncertainty quantification for high-dimensional data.

Findings

CXPlain is more accurate than existing methods.

CXPlain is significantly faster in generating explanations.

Uncertainty estimates correlate with explanation accuracy.

Abstract

Feature importance estimates that inform users about the degree to which given inputs influence the output of a predictive model are crucial for understanding, validating, and interpreting machine-learning models. However, providing fast and accurate estimates of feature importance for high-dimensional data, and quantifying the uncertainty of such estimates remain open challenges. Here, we frame the task of providing explanations for the decisions of machine-learning models as a causal learning task, and train causal explanation (CXPlain) models that learn to estimate to what degree certain inputs cause outputs in another machine-learning model. CXPlain can, once trained, be used to explain the target model in little time, and enables the quantification of the uncertainty associated with its feature importance estimates via bootstrap ensembling. We present experiments that demonstrate that CXPlain is significantly more accurate and faster than existing model-agnostic methods for estimating feature importance. In addition, we confirm that the uncertainty estimates provided by CXPlain ensembles are strongly correlated with their ability to accurately estimate feature importance on held-out data.