Research map / Learning theory and algorithms / Fairness
SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness
Mikhail Yurochkin, Yuekai Sun
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 SenSeI fair regularization instead of ERM | BERT-based toxic comment classifier · Toxic comment detection | CTF score: 0.072±0.002 -> 0.029±0.005 | Improved |
| Use SenSeI fair regularization instead of ERM | BERT-based toxic comment classifier · Toxic comment detection | balanced accuracy: 0.807±0.002 -> 0.791±0.005 | Hurt |
| Use SenSeI fair regularization instead of CLP | BERT-based toxic comment classifier · Toxic comment detection | prediction consistency: 0.763±0.048 -> 0.773±0.043 | Improved |
| Use SenSeI fair regularization instead of SenSR | BERT-based toxic comment classifier · Toxic comment detection | prediction consistency: 0.729±0.044 -> 0.773±0.043 | Improved |
| Use SenSeI fair regularization instead of ERM | BERT-based bios occupation classifier · Bios occupation prediction | balanced accuracy: 0.842±0.002 -> 0.843±0.003 | Improved |
| Use SenSeI fair regularization instead of ERM | BERT-based bios occupation classifier · Bios occupation prediction | CTF score: 0.028±0.001 -> 0.003±0.000 | Improved |
| Use SenSeI fair regularization instead of ERM | Adult income classifier · Adult | Gap_G RMS: .179 -> .043 | Improved |
| Use SenSeI fair regularization instead of ERM | Adult income classifier · Adult | balanced accuracy (%): 82.9 -> 76.8 | Hurt |
| Increase fair regularization strength ρ | SenSeI occupation classifier · Bios occupation prediction | prediction consistency | Improved |
| Use worst-case perturbation search rather than random counterfactual sampling | SenSeI instead of CLP · Toxic comment detection | prediction consistency: 0.763±0.048 -> 0.773±0.043 | Improved |
Later papers that compared with this method
2 later accepted papers report SenSeI as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Learning Antidote Data to Individual Unfairness (ICML 2023 · better on 1 of 1 shared result)
- Post-processing for Individual Fairness (NeurIPS 2021 · better on 3 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Training individually fair ML models with sensitive subspace robustness (ICLR 2020 · similarity 0.92)
- Two Simple Ways to Learn Individual Fairness Metrics from Data (ICML 2020 · similarity 0.90)
- Fairness via Independence: A General Regularization Framework for Machine Learning (ICLR 2026 · similarity 0.89)
- fairret: a Framework for Differentiable Fairness Regularization Terms (ICLR 2024 · similarity 0.88)
- Geometry of Relaxed Fair Regression: A Unified Framework for Aware and Unaware Settings (NeurIPS 2026 · similarity 0.87)
- The Fragility of Fairness: Causal Sensitivity Analysis for Fair Machine Learning (NeurIPS 2024 · similarity 0.87)
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.
