How can I choose an explainer? An Application-grounded Evaluation of Post-hoc Explanations

Abstract

There have been several research works proposing new Explainable AI (XAI) methods designed to generate model explanations having specific properties, or desiderata, such as fidelity, robustness, or human-interpretability. However, explanations are seldom evaluated based on their true practical impact on decision-making tasks. Without that assessment, explanations might be chosen that, in fact, hurt the overall performance of the combined system of ML model + end-users. This study aims to bridge this gap by proposing XAI Test, an application-grounded evaluation methodology tailored to isolate the impact of providing the end-user with different levels of information. We conducted an experiment following XAI Test to evaluate three popular post-hoc explanation methods -- LIME, SHAP, and TreeInterpreter -- on a real-world fraud detection task, with real data, a deployed ML model, and fraud analysts. During the experiment, we gradually increased the information provided to the fraud analysts in three stages: Data Only, i.e., just transaction data without access to model score nor explanations, Data + ML Model Score, and Data + ML Model Score + Explanations. Using strong statistical analysis, we show that, in general, these popular explainers have a worse impact than desired. Some of the conclusion highlights include: i) showing Data Only results in the highest decision accuracy and the slowest decision time among all variants tested, ii) all the explainers improve accuracy over the Data + ML Model Score variant but still result in lower accuracy when compared with Data Only; iii) LIME was the least preferred by users, probably due to its substantially lower variability of explanations from case to case.

arXiv commentAccepted at FAccT'21, the ACM Conference on Fairness, Accountability, and Transparency

References in corpus (17)

• A Unified Approach to Interpreting Model Predictions
• Towards A Rigorous Science of Interpretable Machine Learning
• Learning Important Features Through Propagating Activation Differences
• Interpretable machine learning: definitions, methods, and applications
• Predictive learning via rule ensembles
• Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)
• Towards Robust Interpretability with Self-Explaining Neural Networks
• On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study on Deception Detection
• Explainable AI: Beware of Inmates Running the Asylum Or: How I Learnt to Stop Worrying and Love the Social and Behavioural Sciences
• Distill-and-Compare: Auditing Black-Box Models Using Transparent Model Distillation
• Model Agnostic Supervised Local Explanations
• How do Humans Understand Explanations from Machine Learning Systems? An Evaluation of the Human-Interpretability of Explanation
• Explainable Neural Networks based on Additive Index Models
• Quantifying Interpretability and Trust in Machine Learning Systems
• The Doctor Just Won't Accept That!
• What can AI do for me: Evaluating Machine Learning Interpretations in Cooperative Play
• Case-Based Reasoning for Assisting Domain Experts in Processing Fraud Alerts of Black-Box Machine Learning Models

Cited by in corpus (11)

• Explainable Machine Learning for Public Policy: Use Cases, Gaps, and Research Directions
• Fairness via Explanation Quality: Evaluating Disparities in the Quality of Post hoc Explanations
• The Grammar of Interactive Explanatory Model Analysis
• Considerations When Learning Additive Explanations for Black-Box Models
• What Does Evaluation of Explainable Artificial Intelligence Actually Tell Us? A Case for Compositional and Contextual Validation of XAI Building Blocks
• How Can We Develop Explainable Systems? Insights from a Literature Review and an Interview Study
• Pitfalls of Explainable ML: An Industry Perspective
• Comprehension Is a Double-Edged Sword: Over-Interpreting Unspecified Information in Intelligible Machine Learning Explanations
• A moving target in AI-assisted decision-making: Dataset shift, model updating, and the problem of update opacity
• Deep Natural Language Feature Learning for Interpretable Prediction
• In defence of post-hoc explanations in medical AI