Research map / Federated learning and privacy / Differential privacy
Improving Fairness and Privacy in Selection Problems
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 |
|---|---|---|---|
| Apply the exponential mechanism at the perfect-fairness privacy setting | Qualification-score selection · Synthetic, m=1 | accuracy: 0.96 -> 0.50; Acc. Reduction. 47.92% | Hurt |
| Apply the exponential mechanism at the perfect-fairness privacy setting | Qualification-score selection · Synthetic, m=2 | accuracy: 0.88 -> 0.64; Acc. Reduction. 27.27% | Hurt |
| Apply the exponential mechanism at the perfect-fairness privacy setting | Qualification-score selection · Synthetic, m=3 | accuracy: 0.77 -> 0.73; Acc. Reduction. 5.19% | Hurt |
| Set epsilon to 0 because positive epsilon does not attain perfect fairness | Qualification-score selection · Synthetic, m=4 | accuracy: 0.71 -> 0.4; Acc. Reduction. 44.66% | Hurt |
| Apply the exponential mechanism at the perfect-fairness privacy setting | Qualification-score selection · FICO White-Hispanic vs Asian, m=1 | accuracy: 0.97 -> 0.94; Acc. Reduction. 3.09% | Hurt |
| Apply the exponential mechanism at the perfect-fairness privacy setting | Qualification-score selection · FICO White-Hispanic vs Asian, m=2 | accuracy: 0.96 -> 0.94; Acc. Reduction. 2.10% | Hurt |
| Set epsilon to 0 because positive epsilon does not attain perfect fairness | Qualification-score selection · FICO White-Hispanic vs Asian, m=3 | accuracy: 0.93 -> 0.70; Acc. Reduction. 24.73% | Hurt |
| Set epsilon to 0 because positive epsilon does not attain perfect fairness | Qualification-score selection · FICO White-Hispanic vs Asian, m=4 | accuracy: 0.90 -> 0.70; Acc. Reduction. 22.22% | Hurt |
| Use the exponential mechanism with positive epsilon rather than highest-score selection | Selection among FICO Black and White applicants · FICO Black vs White | equal-opportunity fairness | Helped |
| Use the exponential mechanism with positive epsilon rather than highest-score selection | Selection among FICO White and Black applicants · FICO Black vs White | accuracy | Mixed |
| Use the exponential mechanism with positive epsilon rather than highest-score selection | Selection from a fixed applicant pool · Selection problems | privacy | Helped |
| Increase epsilon | Exponential-mechanism selection · Selection problems under the monotone likelihood ratio assumption | accuracy | Helped |
Closest papers by meaning
- Differentially Private Fair Learning (ICML 2019 · similarity 0.91)
- Fair Sequential Selection Using Supervised Learning Models (NeurIPS 2021 · similarity 0.90)
- Differentially Private Post-Processing for Fair Regression (ICML 2024 · similarity 0.88)
- Differentially Private Model Merging (NeurIPS 2026 · similarity 0.87)
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.86)
- Instance-Specific Asymmetric Sensitivity in Differential Privacy (NeurIPS 2024 · similarity 0.86)
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.