On the Robustness of Global Feature Effect Explanations
arXiv:2406.09069 · doi:10.1007/978-3-031-70344-7_8
Abstract
We study the robustness of global post-hoc explanations for predictive models trained on tabular data. Effects of predictor features in black-box supervised learning are an essential diagnostic tool for model debugging and scientific discovery in applied sciences. However, how vulnerable they are to data and model perturbations remains an open research question. We introduce several theoretical bounds for evaluating the robustness of partial dependence plots and accumulated local effects. Our experimental results with synthetic and real-world datasets quantify the gap between the best and worst-case scenarios of (mis)interpreting machine learning predictions globally.
Accepted at ECML PKDD 2024
References in corpus (9)
- A Comprehensive Taxonomy for Explainable Artificial Intelligence: A Systematic Survey of Surveys on Methods and Concepts
- Adversarial attacks and defenses in explainable artificial intelligence: A survey
- Visualizing the Feature Importance for Black Box Models
- Relating the Partial Dependence Plot and Permutation Feature Importance to the Data Generating Process
- Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations and Beyond
- OpenXAI: Towards a Transparent Evaluation of Model Explanations
- The Grammar of Interactive Explanatory Model Analysis
- Fooling Partial Dependence via Data Poisoning
- iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios