Research map / Learning theory and algorithms / Fairness
Retiring Adult: New Datasets for Fair Machine Learning
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 |
|---|---|---|---|
| Use state-level rather than US-level intervention | GBM with LFR · ACSIncome, US-wide | accuracy: 78.7 \pm 0.1\% -> 79.4 \pm 0.1\% | Helped |
| Use state-level rather than US-level intervention | GBM with LFR · ACSIncome, US-wide | DP violation: 16.6 \pm 0.2\% -> 14.0 \pm 0.2\% | Helped |
| Use state-level rather than US-level intervention | GBM with DP post-processing · ACSIncome, US-wide | accuracy: 79.2 \pm 0.1\% -> 80.2 \pm 0.1\% | Helped |
| Use state-level rather than US-level intervention | GBM with DP post-processing · ACSIncome, US-wide | DP violation: 0.3 \pm 0.3\% -> -0.6 \pm 0.3\% | No clear effect |
| Train and test in a different state instead of the same state | GBM, with or without DP post-processing · ACSIncome, geographic shift across states | accuracy | Mixed |
| Train and test in a different state instead of the same state | GBM with DP post-processing · ACSIncome, geographic shift across states | DP violation | Mixed |
| Evaluate on later years after training on 2014 data | GBM, with or without fairness intervention · ACSIncome, 2014–2018 | accuracy | Hurt |
| Evaluate on later years after training on 2014 data | GBM, with or without fairness intervention · ACSIncome, 2014–2018 | DP violation | No clear effect |
| Increase dataset size and use the newer ACSIncome data rather than IPUMS Adult | Unconstrained GBM · IPUMS Adult 1994 vs ACSIncome 2018 | TPR disparity: 11.5% -> 14.8% | Hurt |
| Vary the income threshold used to define the positive label | GBM and fairness interventions · IPUMS Adult | accuracy | Mixed |
| Vary the income threshold used to define the positive label | Fairness interventions targeting DP or EO · IPUMS Adult | accuracy | Mixed |
| Apply LFR across different state distributions | GBM with LFR · ACSIncome, state-by-state | DP violation: 10\% to close to zero | Mixed |
Closest papers by meaning
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.87)
- Arbitrariness and Social Prediction: The Confounding Role of Variance in Fair Classification (AAAI 2024 · similarity 0.87)
- FairPFN: A Tabular Foundation Model for Causal Fairness (ICML 2025 · similarity 0.86)
- Fairness without Demographics through Shared Latent Space-Based Debiasing (AAAI 2024 · similarity 0.86)
- Size-adaptive Hypothesis Testing for Fairness (NeurIPS 2025 · similarity 0.86)
- Fairness-Accuracy Trade-Offs: A Causal Perspective (AAAI 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.