Research map / Learning theory and algorithms / Fairness
Fairness without Demographics through Adversarially Reweighted Learning
Preethi Lahoti, Alex Beutel, Jilin Chen, Kang Wook Lee, Flavien Prost, Nithum Thain, Xuezhi Wang, Ed Chi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · Adult | AUC (min): 0.867 -> 0.881 | Improved |
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · LSAC | AUC (min): 0.790 -> 0.798 | Improved |
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · COMPAS | AUC (min): 0.674 -> 0.658 | No clear effect |
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · Adult | AUC (minority): 0.875 -> 0.942 | Improved |
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · LSAC | AUC (minority): 0.824 -> 0.832 | Improved |
| Use adversarial reweighting based on computationally identifiable errors | Binary classification without observed demographics · COMPAS | AUC (minority): 0.774 -> 0.785 | Improved |
| Use adversarially learned weights rather than group-based inverse probability weights | Fair classification · LSAC | AUC (min): IPW(S) 0.772; ARL 0.798 | Improved |
| Include class label Y as adversary input (ablation) | ARL with protected features S available · Adult | AUC (min): 0.875 -> 0.882 | Improved |
| Include class label Y as adversary input (ablation) | ARL with protected features S available · LSAC | AUC (min): 0.799 -> 0.801 | Improved |
| Include class label Y as adversary input (ablation) | ARL with protected features S available · COMPAS | AUC (min): 0.675 -> 0.681 | Improved |
| Add protected features S to the adversary input X+Y (ablation) | ARL · Adult | AUC (min): 0.881 -> 0.881 | No clear effect |
| Add protected features S to the adversary input X+Y (ablation) | ARL · LSAC | AUC (min): 0.798 -> 0.808 | Improved |
Closest papers by meaning
- Adversarial Fairness Network (AAAI 2024 · similarity 0.89)
- Fairness without Demographics through Shared Latent Space-Based Debiasing (AAAI 2024 · similarity 0.88)
- On the Fairness ROAD: Robust Optimization for Adversarial Debiasing (ICLR 2024 · similarity 0.88)
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.88)
- Towards Harmless Rawlsian Fairness Regardless of Demographic Prior (NeurIPS 2024 · similarity 0.87)
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 2025 · 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.
