Research map / Adversarial robustness and security / Black-box and transfer attacks
Characterizing the risk of fairwashing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Constrain explainer unfairness to less than 50% of the black-box model's unfairness | Fairwashed logistic regression explainers · Adult Income, Marketing, COMPAS, and Default Credit | fidelity: greater than 92%, 96%, 81% and 90% respectively | Helped |
| Evaluate the fairwashed explainer on data outside the suing group | Fairwashed explanation models · Test set | fidelity-unfairness trade-off | No clear effect |
| Transfer a DNN-teacher fairwashed explainer to other black-box models | Fairwashed logistic regression, Adult Income predictive equality · AdaBoost, RF, and XgBoost students | fidelity: 95% for DNN teacher; 94%, 95% and 93% for AdaBoost, RF and XgBoost | No clear effect |
| Use the range of unfairness in the high-fidelity Rashomon set to assess fairwashing risk | FaiRS analysis of logistic regression explainers · Adult Income versus COMPAS | range of statistical parity | Mixed |
Closest papers by meaning
- Washing The Unwashable : On The (Im)possibility of Fairwashing Detection (NeurIPS 2022 · similarity 0.91)
- Fairwashing: the risk of rationalization (ICML 2019 · similarity 0.91)
- Fooling SHAP with Stealthily Biased Sampling (ICLR 2023 · similarity 0.86)
- An Information-Flow Perspective on Algorithmic Fairness (AAAI 2024 · similarity 0.84)
- Constructing Fair Latent Space for Intersection of Fairness and Explainability (AAAI 2025 · similarity 0.83)
- Procedural Fairness Through Decoupling Objectionable Data Generating Components (ICLR 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.