Getting a CLUE: A Method for Explaining Uncertainty Estimates

Abstract

Both uncertainty estimation and interpretability are important factors for trustworthy machine learning systems. However, there is little work at the intersection of these two areas. We address this gap by proposing a novel method for interpreting uncertainty estimates from differentiable probabilistic models, like Bayesian Neural Networks (BNNs). Our method, Counterfactual Latent Uncertainty Explanations (CLUE), indicates how to change an input, while keeping it on the data manifold, such that a BNN becomes more confident about the input's prediction. We validate CLUE through 1) a novel framework for evaluating counterfactual explanations of uncertainty, 2) a series of ablation experiments, and 3) a user study. Our experiments show that CLUE outperforms baselines and enables practitioners to better understand which input patterns are responsible for predictive uncertainty.

arXiv commentAccepted as an oral presentation at ICLR 2021

References in corpus (19)

• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Towards A Rigorous Science of Interpretable Machine Learning
• Axiomatic Attribution for Deep Networks
• Methods for Interpreting and Understanding Deep Neural Networks
• Counterfactual Fairness
• Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks
• Actionable Recourse in Linear Classification
• Metrics for Explainable AI: Challenges and Prospects
• Bayesian Deep Learning and a Probabilistic Perspective of Generalization
• Practical Deep Learning with Bayesian Principles
• Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning
• From Parity to Preference-based Notions of Fairness in Classification
• Explaining Image Classifiers by Counterfactual Generation
• Estimating or Propagating Gradients Through Stochastic Neurons
• Distribution Matching in Variational Inference
• Uncertainty as a Form of Transparency: Measuring, Communicating, and Using Uncertainty
• xGEMs: Generating Examplars to Explain Black-Box Models
• Disentangling and Learning Robust Representations with Natural Clustering
• Avoiding Resentment Via Monotonic Fairness

Cited by in corpus (14)

• Towards Human-centered Explainable AI: A Survey of User Studies for Model Explanations
• Machine Learning Explainability for External Stakeholders
• Sample-Efficient Optimization in the Latent Space of Deep Generative Models via Weighted Retraining
• Calibrated Explanations: with Uncertainty Information and Counterfactuals
• DS-UI: Dual-Supervised Mixture of Gaussian Mixture Models for Uncertainty Inference
• Explaining Predictive Uncertainty by Exposing Second-Order Effects
• Explaining in Style: Training a GAN to explain a classifier in StyleSpace
• To Trust or Not to Trust a Regressor: Estimating and Explaining Trustworthiness of Regression Predictions
• Endogenous Macrodynamics in Algorithmic Recourse
• A Trustworthiness Score to Evaluate DNN Predictions
• Enhancing Counterfactual Explanation Search with Diffusion Distance and Directional Coherence
• Shapley variable importance clouds for interpretable machine learning
• Towards User Guided Actionable Recourse
• Scene Uncertainty and the Wellington Posterior of Deterministic Image Classifiers