CoChat
Sign Up
Cited by 28Ethics and Social Impacts of AIMachine Learning and Data ClassificationPrivacy-Preserving Technologies in Data

Promoting Fairness through Hyperparameter Optimization

Open access
F.Cruz, Andre, Saleiro, Pedro, Belem, Catarina, Soares, Carlos, Bizarro, Pedro·2021·2021 IEEE International Conference on Data Mining (ICDM)
DOI 10.1109/icdm51629.2021.00119
Open this paper in CoChat — free accountOpen PDF

Abstract

Considerable research effort has been guided towards algorithmic fairness but real-world adoption of bias reduction techniques is still scarce. Existing methods are either metric-or model-specific, require access to sensitive attributes at inference time, or carry high development or deployment costs. This work explores the unfairness that emerges when optimizing ML models solely for predictive performance, and how to mitigate it with a simple and easily deployed intervention: fairness-aware hyperparameter optimization (HO). We propose and evaluate fairness-aware variants of three popular HO algorithms: Fair Random Search, Fair TPE, and Fairband. We validate our approach on a real-world bank account opening fraud case-study, as well as on three datasets from the fairness literature. Results show that, without extra training cost, it is feasible to find models with 111% mean fairness increase and just 6% decrease in performance when compared with fairness-blind HO.1

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/icdm51629.2021.00119

Cites

14 of 14 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Prior studies have shown that prioritising fairness-unaware model optimisation-e.g. by focusing solely on the model's accuracy-can lead to inadvertently prioritising models with higher levels of bias.

    “Unfortunately, as prior studies have shown [34,120,121,156], prioritising such fairness-unaware model optimisation-e.g. by focusing solely on the model's accuracy-can lead to inadvertently prioritising models with higher levels of bias.”

    From Out of Context: Investigating the Bias and Fairness Concerns of “Artificial Intelligence as a Service” (2023) · cites this paper as [34]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 22

  1. [3][3] Joy Buolamwini and Timnit Gebru. Gender shades: Intersectional accuracy disparities in commercial gender classification. In FAT* '18 , volume 81, pages 77-91. PMLR, 23-24 Feb 2018. -
  2. [5][5] Kit T Rodolfa, Pedro Saleiro, and Rayid Ghani. Chapter 11: Bias and Fairness. In Big data and social science: A practical guide to methods and tools . CRC Press, 2020. -
  3. [6][6] Muhammad Bilal Zafar, Isabel Valera, Manuel Gomez Rogriguez, and Krishna P. Gummadi. Fairness Constraints: Mechanisms for Fair Classification. In 20th AISTATS , volume 54, pages 962-970. PMLR, 20-22 Apr 2017. -
  4. [7][7] Alekh Agarwal, Alina Beygelzimer, Miroslav Dudik, John Langford, and Hanna Wallach. A Reductions Approach to Fair Classification. In 35th ICML , volume 80, pages 60-69. PMLR, 2018. -
  5. [11][11] European Parliament and Council. Directive 2014/92/EU of the European Parliament and of the Council of 23 July 2014 on the comparability of fees related to payment accounts, payment account switching and access to payment accounts with basic features. Official Journal of the European Union (OJ) , 57:214, 2014. -
  6. [12][12] Flavio Calmon, Dennis Wei, Bhanukiran Vinzamuri, Karthikeyan Natesan Ramamurthy, and Kush R Varshney. Optimized pre-processing for discrimination prevention. In NIPS , pages 3992-4001, 2017. -
  7. [13][13] Elliot Creager, David Madras, J¨ orn-Henrik Henrik Jacobsen, Marissa A. Weis, Kevin Swersky, Toniann Pitassi, and Richard Zemel. Flexibly Fair Representation Learning by Disentanglement. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors, Proc. of the 36th ICML , volume 2019-June, pages 2600-2612. PMLR, jun 2019. -
  8. [14][14] Harrison Edwards and Amos Storkey. Censoring Representations with an Adversary. In 4th Int. Conf. on Learning Representations, ICLR 2016 - Conf. Track Proceedings , pages 1-14, 2016. -
  9. [15][15] Richard S Zemel, Yu Wu, Kevin Swersky, Toniann Pitassi, and Cynthia Dwork. Learning fair representations. 30th ICML , 28(PART 2):325-333, 2013. -
  10. [17][17] Benjamin Fish, Jeremy Kun, and ´ Ad´ am D. Lelkes. A confidence-based approach for balancing fairness and accuracy. Proc. of the 2016 SIAM Int. Conf. on Data Mining , pages 144-152, 2016. -
  11. [18][18] Moritz Hardt, Eric Price, Eric Price, and Nati Srebro. Equality of opportunity in supervised learning. In NIPS , pages 3315-3323. Curran Associates, Inc., 2016. -
  12. [20][20] James Bergstra and Yoshua Bengio. Random search for hyper-parameter optimization. JMLR , 13:281-305, 2012. -
  13. [21][21] Bobak Shahriari, Kevin Swersky, Ziyu Wang, Ryan P. Adams, and Nando De Freitas. Taking the human out of the loop: A review of Bayesian optimization. Proc. of the IEEE , 104(1):148-175, 2016. -
  14. [23][23] Michael A. Gelbart, Jasper Snoek, and Ryan P. Adams. Bayesian optimization with unknown constraints. In 30th UAI , pages 250-259. AUAI Press, 2014. -
  15. [26][26] Emanuel Parzen. On estimation of a probability density function and mode. Ann. Math. Statist. , 33(3):1065-1076, 09 1962. -
  16. [28][28] Zohar Karnin, Tomer Koren, and Oren Somekh. Almost optimal exploration in multi-armed bandits. 30th ICML , 28:2275-2283, 2013. -
  17. [30][30] Jean-Yves Audibert and S´ ebastien Bubeck. Best Arm Identification in Multi-Armed Bandits. In COLT - 23th Conf. on Learning Theory - 2010 , page 13 p., Haifa, Israel, jun 2010. -
  18. [34][34] O. Schutze, A. Lara, and C. A. C. Coello. On the influence of the number of objectives on the hardness of a multiobjective optimization problem. IEEE TEC , 15(4):444-455, 2011. -
  19. [36][36] Sam Corbett-Davies, Emma Pierson, Avi Feller, Sharad Goel, and Aziz Huq. Algorithmic Decision Making and the Cost of Fairness. In 23rd ACM SIGKDD , pages 797-806. ACM Press, jan 2017. -
  20. [39][39] Aisha Abdallah, Mohd Aizaini Maarof, and Anazida Zainal. Fraud detection system: A survey. Journal of Network and Computer Applications , 68:90 - 113, 2016. -
  21. [41][41] Leo Breiman. Random Forests. Machine Learning , 45(1):1-33, 2001. -
  22. [42][42] Leo Breiman, Jerome Friedman, Charles J Stone, and Richard A Olshen. Classification and regression trees . CRC press, 1984. -

Cited by

Sorted by citations

Related

Authors
F.Cruz, A., Saleiro, P., Belem, C., Soares, C., & Bizarro, P.
Affiliations
Universidade do Porto (PT)
ORCID
Venue
2021 IEEE International Conference on Data Mining (ICDM)
Publisher
IEEE
Published
2021-12
Pages
1036-1041
ISSN
1550-4786 (ISSN-L)
DOI
10.1109/icdm51629.2021.00119
OpenAlex
W4206958560
Cited by
28
References
14
Citations by year
2026: 8 · 2025: 3 · 2024: 8 · 2023: 8 · 2021: 1
FWCI
3.84
Type
conference paper
Language
en
Open access
Licence
Locations
  • Proceedings (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
  • Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
Keywords
algorithmic fairness, hyperparameter optimization, bias mitigation, deployment cost

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review