mlsanityData to 7 Oct 2026

Multiaccurate Proxies for Downstream Fairness

arXiv 2107.04423 ↗First posted 9 Jul 2021cs.LG cs.DScited by 17

Not on the map. Why?

Abstract (arXiv)

We study the problem of training a model that must obey demographic fairness conditions when the sensitive features are not available at training time -- in other words, how can we train a model to be fair by race when we don't have data about race? We adopt a fairness pipeline perspective, in which an "upstream" learner that does have access to the sensitive features will learn a proxy model for these features from the other attributes. The goal of the proxy is to allow a general "downstream" learner -- with minimal assumptions on their prediction task -- to be able to use the proxy to train a model that is fair with respect to the true sensitive features. We show that obeying multiaccuracy constraints with respect to the downstream model class suffices for this purpose, provide sample- and oracle efficient-algorithms and generalization bounds for learning such proxies, and conduct an experimental evaluation. In general, multiaccuracy is much easier to satisfy than classification accuracy, and can be satisfied even when the sensitive features are hard to predict.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Builds on 19 of its 41 references that are in the map. First posted on arXiv 9 Jul 2021.

Builds on

19 of 41 references.

References · 19

  1. 2016Fair prediction with disparate impact: A study of bias in recidivism prediction instrumentscited by 671
  2. 2018A Reductions Approach to Fair Classificationcited by 585
  3. 2017Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairnesscited by 381
  4. 2020Fairness without Demographics through Adversarially Reweighted Learningcited by 251
  5. 2018Multiaccuracy: Black-Box Post-Processing for Fairness in Classificationcited by 221
  6. 2020Minimax Pareto Fairness: A Multi Objective Perspectivecited by 152
  7. 2018Fairness Under Unawareness: Assessing Disparity When Protected Class Is Unobservedcited by 94
  8. 2019Assessing Algorithmic Fairness with Unobserved Protected Class Using Data Combinationcited by 79
  9. 2020Robust Optimization for Fairness with Noisy Protected Groupscited by 78
  10. 2018Differentially Private Fair Learningcited by 76
  11. 2020Moment Multicalibration for Uncertainty Estimationcited by 63
  12. 2019Equalized odds postprocessing under imperfect group informationcited by 48
7 more
  1. 2018Blind Justice: Fairness with Encrypted Sensitive Attributescited by 41
  2. 2018Downstream Effects of Affirmative Actioncited by 34
  3. 2021Evaluating Fairness of Machine Learning Models Under Uncertain and Incomplete Informationcited by 33
  4. 2020Mitigating Bias in Set Selection with Noisy Protected Attributescited by 29
  5. 2020Individual Fairness in Pipelinescited by 22
  6. 2017Fair Pipelinescited by 21
  7. 2021Lexicographically Fair Learning: Algorithms and Generalizationcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 2; 2022: 7; 2023: 4; 2024: 2; 2025: 1; 2026: 1.

17 papers.

Sort:

  1. 2022Bias Mitigation for Machine Learning Classifiers: A Comprehensive Surveycited by 125
  2. 2021The Rich Get Richer: Disparate Impact of Semi-Supervised Learningcited by 28
  3. 2021Simple and near-optimal algorithms for hidden stratification and multi-group learningcited by 26
  4. 2023Ground(less) Truth: A Causal Framework for Proxy Labels in Human-Algorithm Decision-Makingcited by 25
  5. 2022Weak Proxies are Sufficient and Preferable for Fairness with Missing Sensitive Attributescited by 17
  6. 2022Multicalibrated Regression for Downstream Fairnesscited by 13
  7. 2022Comparative Learning: A Sample Complexity Theory for Two Hypothesis Classescited by 8
  8. 2024Fairness without Sensitive Attributes via Knowledge Sharingcited by 6
  9. 2023Fairness Under Demographic Scarce Regimecited by 4
  10. 2023Toward the Tradeoffs between Privacy, Fairness and Utility in Federated Learningcited by 4
  11. 2022Context matters for fairness -- a case study on the effect of spatial distribution shiftscited by 3
  12. 2022"You Can't Fix What You Can't Measure": Privately Measuring Demographic Performance Disparities in Federated Learningcited by 3
5 more
  1. 2022Distributionally Robust Data Joincited by 2
  2. 2025Multiaccuracy and Multicalibration via Proxy Groupscited by 1
  3. 2026Safe Fairness Guarantees Without Demographics in Classification: Spectral Uncertainty Set Perspectivecited by 1
  4. 2023Balanced Filtering via Disclosure-Controlled Proxiescited by 0
  5. 2024Dancing in the Shadows: Harnessing Ambiguity for Fairer Classifierscited by 0