CXPlain: Causal Explanations for Model Interpretation under Uncertainty

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.

To appear in Advances in Neural Information Processing Systems 2019

Cited by in corpus (11)

• CASTLE: Regularization via Auxiliary Causal Graph Discovery
• Self-Explaining Structures Improve NLP Models
• Adversarial Infidelity Learning for Model Interpretation
• CDT: Cascading Decision Trees for Explainable Reinforcement Learning
• Pair the Dots: Jointly Examining Training History and Test Stimuli for Model Interpretability
• Have We Learned to Explain?: How Interpretability Methods Can Learn to Encode Predictions in their Interpretations
• Generative Interventions for Causal Learning
• Shapley variable importance clouds for interpretable machine learning
• Neural-to-Tree Policy Distillation with Policy Improvement Criterion
• Visualizing Color-wise Saliency of Black-Box Image Classification Models
• Influence-Driven Explanations for Bayesian Network Classifiers