Research map / Learning theory and algorithms / Fairness
Path-Specific Counterfactual Fairness
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 |
|---|---|---|---|
| Correct unfair direct and mediated effects using latent inference-projection | Berkeley Admission · Biased Berkeley Admission test set | accuracy: 71.6% -> 67.1% | Hurt |
| Use department choice only, omitting the direct sensitive-attribute link | Berkeley Admission · Biased Berkeley Admission | accuracy: 71.6% -> 67.9% | Hurt |
| Constrain PSE between 0.95 and 1.05 instead of leaving the model unconstrained | UCI Adult · UCI Adult test set | accuracy: 82% -> 72% | Hurt |
| Apply counterfactual correction for both sexes (ablation) | UCI Adult latent inference-projection · UCI Adult test set, 5,000 steps | fair accuracy: 82.80% -> 81.66% | Hurt |
| Apply counterfactual correction for both sexes (ablation) | UCI Adult latent inference-projection · UCI Adult test set, 8,000 steps | fair accuracy: 80.05% -> 80.21% | Helped |
| Apply counterfactual correction for both sexes (ablation) | UCI Adult latent inference-projection · UCI Adult test set, 15,000 steps | fair accuracy: 80.87% -> 79.41% | Hurt |
| Apply counterfactual correction for both sexes (ablation) | UCI Adult latent inference-projection · UCI Adult test set, 20,000 steps | fair accuracy: 76.44% -> 73.98% | Hurt |
| Correct unfair sensitive-attribute pathways | UCI German Credit latent inference-projection · UCI German Credit test set | accuracy: 74.67% -> 73.67%; 76.33% -> 76.00%; 76.00% -> 76.00% | No clear effect |
| Increase training with MMD penalization from 5,000 to 15,000 steps | UCI Adult latent inference-projection · UCI Adult test set | fair accuracy: 81.66% -> 79.41% | Hurt |
| Increase training with MMD penalization to 20,000 steps | UCI Adult latent inference-projection · UCI Adult test set | fair accuracy: 79.41% -> 73.98% | Hurt |
Closest papers by meaning
- Counterfactual Fairness by Combining Factual and Counterfactual Predictions (NeurIPS 2024 · similarity 0.88)
- Fair Off-Policy Learning from Observational Data (ICML 2024 · similarity 0.86)
- Fairness on Principal Stratum: A New Perspective on Counterfactual Fairness (ICML 2025 · similarity 0.86)
- Fairness-Accuracy Trade-Offs: A Causal Perspective (AAAI 2025 · similarity 0.85)
- Towards counterfactual fairness through auxiliary variables (ICLR 2025 · similarity 0.85)
- FairPFN: A Tabular Foundation Model for Causal Fairness (ICML 2025 · similarity 0.85)
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.