Explaining Machine Learning Classifiers through Diverse Counterfactual Explanations

Abstract

Post-hoc explanations of machine learning models are crucial for people to understand and act on algorithmic predictions. An intriguing class of explanations is through counterfactuals, hypothetical examples that show people how to obtain a different prediction. We posit that effective counterfactual explanations should satisfy two properties: feasibility of the counterfactual actions given user context and constraints, and diversity among the counterfactuals presented. To this end, we propose a framework for generating and evaluating a diverse set of counterfactual explanations based on determinantal point processes. To evaluate the actionability of counterfactuals, we provide metrics that enable comparison of counterfactual-based methods to other local explanation methods. We further address necessary tradeoffs and point to causal implications in optimizing for counterfactuals. Our experiments on four real-world datasets show that our framework can generate a set of counterfactuals that are diverse and well approximate local decision boundaries, outperforming prior approaches to generating diverse counterfactuals. We provide an implementation of the framework at https://github.com/microsoft/DiCE.

13 pages

References in corpus (6)

• Adam: A Method for Stochastic Optimization
• Towards A Rigorous Science of Interpretable Machine Learning
• Counterfactual Fairness
• Determinantal point processes for machine learning
• Local Rule-Based Explanations of Black Box Decision Systems
• Interpretable to Whom? A Role-based Model for Analyzing Interpretable Machine Learning Systems

Cited by in corpus (151)

• Interpretable Machine Learning -- A Brief History, State-of-the-Art and Challenges
• The Hidden Assumptions Behind Counterfactual Explanations and Principal Reasons
• Contrastive Explanation: A Structural-Model Approach
• A Prescriptive Learning Analytics Framework: Beyond Predictive Modelling and onto Explainable AI with Prescriptive Analytics and ChatGPT
• Preserving Causal Constraints in Counterfactual Explanations for Machine Learning Classifiers
• Counterfactual Explanations for Machine Learning on Multivariate Time Series Data
• Towards Unifying Feature Attribution and Counterfactual Explanations: Different Means to the Same End
• Algorithmic recourse under imperfect causal knowledge: a probabilistic approach
• Algorithmic Recourse: from Counterfactual Explanations to Interventions
• Explainable Machine Learning for Public Policy: Use Cases, Gaps, and Research Directions
• Understanding the Effect of Counterfactual Explanations on Trust and Reliance on AI for Human-AI Collaborative Clinical Decision Making
• Counterfactual State Explanations for Reinforcement Learning Agents via Generative Deep Learning
• Counterfactual Shapley Additive Explanations
• Towards Relatable Explainable AI with the Perceptual Process
• Counterfactual Explanations for Neural Recommenders
• Explainable Fairness in Recommendation
• A survey of algorithmic recourse: definitions, formulations, solutions, and prospects
• Directive Explanations for Monitoring the Risk of Diabetes Onset: Introducing Directive Data-Centric Explanations and Combinations to Support What-If Explorations
• Causality Learning: A New Perspective for Interpretable Machine Learning
• Machine Learning for Observational Cosmology
• A comprehensive interpretable machine learning framework for Mild Cognitive Impairment and Alzheimer's disease diagnosis
• A Framework and Benchmarking Study for Counterfactual Generating Methods on Tabular Data
• Explainable Machine Learning with Prior Knowledge: An Overview
• Model-Based Counterfactual Synthesizer for Interpretation
• Scientific Inference With Interpretable Machine Learning: Analyzing Models to Learn About Real-World Phenomena
• ReLAX: Reinforcement Learning Agent eXplainer for Arbitrary Predictive Models
• Improving Visualization Interpretation Using Counterfactuals
• Beyond Individualized Recourse: Interpretable and Interactive Summaries of Actionable Recourses
• Sparse Visual Counterfactual Explanations in Image Space
• Model extraction from counterfactual explanations
• Rational Shapley Values
• On Counterfactual Explanations under Predictive Multiplicity
• Counterfactuals and Causability in Explainable Artificial Intelligence: Theory, Algorithms, and Applications
• GAM Coach: Towards Interactive and User-centered Algorithmic Recourse
• Calibrated Explanations: with Uncertainty Information and Counterfactuals
• Benchmarking Instance-Centric Counterfactual Algorithms for XAI: From White Box to Black Box
• Consequence-aware Sequential Counterfactual Generation
• From Explanation to Recommendation: Ethical Standards for Algorithmic Recourse
• Unraveling the Dilemma of AI Errors: Exploring the Effectiveness of Human and Machine Explanations for Large Language Models
• Human-Centered Explainable AI (XAI): From Algorithms to User Experiences
• ViCE: Visual Counterfactual Explanations for Machine Learning Models
• Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyond
• Generative causal explanations of black-box classifiers
• Achieving Diversity in Counterfactual Explanations: a Review and Discussion
• Evolutionary approaches to explainable machine learning
• Latent-CF: A Simple Baseline for Reverse Counterfactual Explanations
• Supporting Better Insights of Data Science Pipelines with Fine-grained Provenance
• Detecting Violations of Differential Privacy for Quantum Algorithms
• Decisions, Counterfactual Explanations and Strategic Behavior
• Beyond Model Interpretability: Socio-Structural Explanations in Machine Learning
• GeCo: Quality Counterfactual Explanations in Real Time
• The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models
• Causal Interpretability for Machine Learning -- Problems, Methods and Evaluation
• From Black Boxes to Conversations: Incorporating XAI in a Conversational Agent
• Towards Interpretability in Audio and Visual Affective Machine Learning: A Review
• Synthesizing explainable counterfactual policies for algorithmic recourse with program synthesis
• DECE: Decision Explorer with Counterfactual Explanations for Machine Learning Models
• Explaining Visual Models by Causal Attribution
• CARLA: A Python Library to Benchmark Algorithmic Recourse and Counterfactual Explanation Algorithms
• Ensemble of Counterfactual Explainers
• Counterfactual Explanations for Arbitrary Regression Models
• Explaining in Style: Training a GAN to explain a classifier in StyleSpace
• Benign Shortcut for Debiasing: Fair Visual Recognition via Intervention with Shortcut Features
• Feature-based Learning for Diverse and Privacy-Preserving Counterfactual Explanations
• GradCFA: A Hybrid Gradient-Based Counterfactual and Feature Attribution Explanation Algorithm for Local Interpretation of Neural Networks
• TABCF: Counterfactual Explanations for Tabular Data Using a Transformer-Based VAE
• Learning Global Transparent Models Consistent with Local Contrastive Explanations
• CounteRGAN: Generating Realistic Counterfactuals with Residual Generative Adversarial Nets
• Algorithmic Recourse in the Wild: Understanding the Impact of Data and Model Shifts
• Optimal Counterfactual Explanations in Tree Ensembles
• Explaining AI Decisions: Towards Achieving Human-Centered Explainability in Smart Home Environments
• Memory Wrap: a Data-Efficient and Interpretable Extension to Image Classification Models
• "Is It My Turn?" Assessing Teamwork and Taskwork in Collaborative Immersive Analytics
• Consistent Counterfactuals for Deep Models
• Learning Fair Rule Lists
• On the Connection between Game-Theoretic Feature Attributions and Counterfactual Explanations
• RelEx: A Model-Agnostic Relational Model Explainer
• LLM-Powered Prediction of Hyperglycemia and Discovery of Behavioral Treatment Pathways from Wearables and Diet
• Conditional Generative Models for Counterfactual Explanations
• Calibrated Explanations for Regression
• RoCourseNet: Distributionally Robust Training of a Prediction Aware Recourse Model
• The Intriguing Relation Between Counterfactual Explanations and Adversarial Examples
• Guiding the generation of counterfactual explanations through temporal background knowledge for Predictive Process Monitoring
• Counterfactual Evaluation for Explainable AI
• Optimal Counterfactual Explanations for Scorecard modelling
• Endogenous Macrodynamics in Algorithmic Recourse
• An exact counterfactual-example-based approach to tree-ensemble models interpretability
• Challenging common interpretability assumptions in feature attribution explanations
• For Better or Worse: The Impact of Counterfactual Explanations' Directionality on User Behavior in xAI
• Ordered Counterfactual Explanation by Mixed-Integer Linear Optimization
• The Out-of-Distribution Problem in Explainability and Search Methods for Feature Importance Explanations
• A unified framework for evaluating the robustness of machine-learning interpretability for prospect risking
• Counterfactual Explanations in Sequential Decision Making Under Uncertainty
• Enhancing Counterfactual Explanation Search with Diffusion Distance and Directional Coherence
• Abstraction Alignment: Comparing Model-Learned and Human-Encoded Conceptual Relationships
• Feature Attributions and Counterfactual Explanations Can Be Manipulated
• Even-Ifs From If-Onlys: Are the Best Semi-Factual Explanations Found Using Counterfactuals As Guides?
• On Quantitative Evaluations of Counterfactuals
• DISSECT: Disentangled Simultaneous Explanations via Concept Traversals
• Framing Algorithmic Recourse for Anomaly Detection
• A framework for predicting, interpreting, and improving Learning Outcomes
• QUCE: The Minimisation and Quantification of Path-Based Uncertainty for Generative Counterfactual Explanations
• A Review of Explainable Artificial Intelligence in Manufacturing
• The Power of Typed Affine Decision Structures: A Case Study
• Towards Human Cognition Level-based Experiment Design for Counterfactual Explanations (XAI)
• Survey on AI Ethics: A Socio-technical Perspective
• Flexible and Robust Counterfactual Explanations with Minimal Satisfiable Perturbations
• Counterfactual Explanations Can Be Manipulated
• Refining the Structure of Neural Networks Using Matrix Conditioning
• Model-agnostic and Scalable Counterfactual Explanations via Reinforcement Learning
• Global Counterfactual Directions
• Seven challenges for harmonizing explainability requirements
• FastIF: Scalable Influence Functions for Efficient Model Interpretation and Debugging
• Causal Inference with Deep Causal Graphs
• Improving Fair Predictions Using Variational Inference In Causal Models
• Towards a multi-stakeholder value-based assessment framework for algorithmic systems
• Navigating Explanatory Multiverse Through Counterfactual Path Geometry
• Towards User Guided Actionable Recourse
• Using Constraints to Discover Sparse and Alternative Subgroup Descriptions
• Counterfactual Explanations with Probabilistic Guarantees on their Robustness to Model Change
• Towards Feasible Counterfactual Explanations: A Taxonomy Guided Template-based NLG Method
• Explaining Black-Box Models through Counterfactuals
• SIBILA: A novel interpretable ensemble of general-purpose machine learning models applied to medical contexts
• A Fully Bayesian, Logistic Regression Tracking Algorithm for Mitigating Disparate Misclassification
• Ca2+ transient detection and segmentation with the Astronomically motivated algorithm for Background Estimation And Transient Segmentation (Astro-BEATS)
• The Impact of Machine Learning Uncertainty on the Robustness of Counterfactual Explanations
• Finding Optimal Diverse Feature Sets with Alternative Feature Selection
• SafeAR: Safe Algorithmic Recourse by Risk-Aware Policies
• Minimal Sufficient Views: A DNN model making predictions with more evidence has higher accuracy
• Discriminative Attribution from Counterfactuals
• Perfect Counterfactuals in Imperfect Worlds: Modelling Noisy Implementation of Actions in Sequential Algorithmic Recourse
• CounterNet: End-to-End Training of Prediction Aware Counterfactual Explanations
• MCCE: Monte Carlo sampling of realistic counterfactual explanations
• Augmenting BERT Carefully with Underrepresented Linguistic Features
• DA-DGCEx: Ensuring Validity of Deep Guided Counterfactual Explanations With Distribution-Aware Autoencoder Loss
• A Series of Unfortunate Counterfactual Events: the Role of Time in Counterfactual Explanations
• Counterfactual Explanation with Multi-Agent Reinforcement Learning for Drug Target Prediction
• How can we trust opaque systems? Criteria for robust explanations in XAI
• Evaluation of AI Ethics Tools in Language Models: A Developers' Perspective Case Study
• Post-discovery Analysis of Anomalous Subsets
• Uncovering Discrimination Clusters: Quantifying and Explaining Systematic Fairness Violations
• Navigating Time's Possibilities: Plausible Counterfactual Explanations for Multivariate Time-Series Forecast through Genetic Algorithms
• Causal Explanations of Image Misclassifications
• Explainability Assistant: A Conversational XAI Interface for Interpreting Energy Consumption Models
• Realistic Counterfactual Explanations via Denial Constraints
• Self-learn to Explain Siamese Networks Robustly
• Learning Models for Actionable Recourse
• Low-Cost Algorithmic Recourse for Users With Uncertain Cost Functions
• Counterfactual Explanations as Plans
• Comprehensible Counterfactual Explanation on Kolmogorov-Smirnov Test
• Towards Designing Computer Vision-based Explainable-AI Solution: A Use Case of Livestock Mart Industry