School of Computer Science, University College Dublin, Dublin 4, Ireland E-mail: william.blanzeisky@ucdconnect.ie, padraig.cunningham@ucd.ie"/>
Search
  • All Journals
Advanced Search
Search articles by Title, Author, Keywords, DOI
Submit Manuscript
7.7CiteScore
2.0Impact Factor
Advanced Search
MAP  >  The Knowledge Engineering Review
2022 Volume 37
Article Contents
Next Previous
RESEARCH ARTICLE   Open Access    

Using Pareto simulated annealing to address algorithmic bias in machine learning

More Information
  • Abstract: Algorithmic bias arises in machine learning when models that may have reasonable overall accuracy are biased in favor of ‘good’ outcomes for one side of a sensitive category, for example gender or race. The bias will manifest as an underestimation of good outcomes for the under-represented minority. In a sense, we should not be surprised that a model might be biased when it has not been ‘asked’ not to be; reasonable accuracy can be achieved by ignoring the under-represented minority. A common strategy to address this issue is to include fairness as a component in the learning objective. In this paper, we consider including fairness as an additional criterion in model training and propose a multi-objective optimization strategy using Pareto Simulated Annealing that optimizes for both accuracy and underestimation bias. Our experiments show that this strategy can identify families of models with members representing different accuracy/fairness tradeoffs. We demonstrate the effectiveness of this strategy on two synthetic and two real-world datasets.
  • 加载中
  • Amine , K. 2019. Multiobjective simulated annealing: Principles and algorithm variants. Advances in Operations Research 2019, 1–13.

    Google Scholar

    Baeza-Yates , R. 2018. Bias on the web. Communications of the ACM 61(6), 54–61.

    Google Scholar

    Blanzeisky , W. & Cunningham , P. 2021. Algorithmic factors influencing bias in machine learning. ECML-PKDD 2021 workshop on Bias and Fairness in AI (arXiv preprint arXiv:2104.14014).

    Google Scholar

    Blanzeisky , W., Cunningham , P. & Kennedy , K. 2021. Introducing a family of synthetic datasets for research on bias in machine learning. arXiv preprint arXiv:2107.08928.

    Google Scholar

    Caton , S. & Haas , C. 2020. Fairness in machine learning: A survey. arXiv preprint arXiv:2010.04053.

    Google Scholar

    Cotter , A., Friedlander , M. P., Goh , G. & Gupta , M. R. 2016. Satisfying real-world goals with dataset constraints. CoRR abs/1606.07558.

    Google Scholar

    Cunha , M. & Marques , J. 2020. A new multiobjective simulated annealing algorithmmosa-gr: Application to the optimal design of water distribution networks. Water Resources Research 56(3), e2019WR025852.

    Google Scholar

    Cunningham , P. & Delany , S. J. 2020. Underestimation bias and underfitting in machine learning. In Trustworthy AI - Integrating Learning, Optimization and Reasoning - First International Workshop, TAILOR 2020, Virtual Event, September 4–5, 2020, Revised Selected Papers, F. Heintz , M. Milano & B. O’Sullivan (eds). Lecture Notes in Computer Science. 12641, 20–31. Springer. https://doi.org/10.1007/978-3-030-73959-1_2

    Google Scholar

    Czyzżak , P. & Jaszkiewicz , A. 1998. Pareto simulated annealing-a metaheuristic technique for multiple-objective combinatorial optimization. Journal of Multi-Criteria Decision Analysis 7(1), 34–47.

    Google Scholar

    Deb , K. 2014. Multi-objective Optimization, 403–449.

    Google Scholar

    Dressel , J. & Farid , H. 2018. The accuracy, fairness, and limits of predicting recidivism. Science Advances 4(1), eaao5580.

    Google Scholar

    Dutta , S., Wei , D., Yueksel , H., Chen , P.-Y., Liu , S. & Varshney , K. R. 2020. Is There a Trade-Off between Fairness and Accuracy? A Perspective using Mismatched Hypothesis Testing.

    Google Scholar

    Feldman , M., Friedler , S. A., Moeller , J., Scheidegger , C. & Venkatasubramanian , S. 2015. Certifying and removing disparate impact. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 259–268.

    Google Scholar

    Foley , K., Greene , D. & Cunningham , P. 2010. Optimizing conflicting objectives in nmf using pareto simulated annealing. In Paper presented at the 21st National Conference on Artificial Intelligence and Cognitive Science (AICS 2010), 112–121.

    Google Scholar

    Goel , N., Yaghini , M. & Faltings , B. 2018. Non-discriminatory machine learning through convex fairness criteria.

    Google Scholar

    Hooker , S., Moorosi , N., Clark , G., Bengio , S. & Denton , E. 2020. Characterising bias in compressed models. arXiv e-prints pp. arXiv–2010.

    Google Scholar

    Japkowicz , N. & Stephen , S. 2002. The class imbalance problem: A systematic study. Intelligent Data Analysis 6(5), 429–449.

    Google Scholar

    Jaszkiewicz , A. 2001. Multiple objective metaheuristic algorithms for combinatorial optimization.

    Google Scholar

    Kamishima , T., Akaho , S., Asoh , H. & Sakuma , J. 2012. Fairness-aware classifier with prejudice remover regularizer. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 35–50. Springer.

    Google Scholar

    Kirkpatrick , S., Gelatt , C. D. & Vecchi , M. P. 1983. Optimization by simulated annealing. Science 220(4598), 671–680.

    Google Scholar

    Kohavi , R. 1996. Scaling up the accuracy of Naive-Bayes classifiers: A decision-tree hybrid. In Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, 96, 202–207.

    Google Scholar

    Mas-Colell , A., Whinston , M. D., Green , J. R. et al. 1995. Microeconomic Theory, 1. Oxford University Press.

    Google Scholar

    Pllana , S., Memeti , S. & Kolodziej , J. 2019. Customizing Pareto simulated annealing for multi-objective optimization of control cabinet layout. In 2019 22nd International Conference on Control Systems and Computer Science (CSCS), 78–85.

    Google Scholar

    Zafar , M. B., Valera , I., Gomez Rodriguez , M. & Gummadi , K. P. 2017. Fairness beyond disparate treatment & disparate impact: Learning classification without disparate mistreatment. In Proceedings of the 26th International Conference on World Wide Web, 1171–1180.

    Google Scholar

    Zafar , M., Valera , I., Rodriguez , M. & Gummadi , K. P. 2015. Fairness Constraints: A Mechanism for Fair Classification.

    Google Scholar

    Zemel , R., Wu , Y., Swersky , K., Pitassi , T. & Dwork , C. 2013. Learning fair representations. In Proceedings of the 30th International Conference on Machine Learning, S. Dasgupta & D. McAllester (eds). Proceedings of Machine Learning Research, 28, 325–333. PMLR.

    Google Scholar

    Zliobaite , I. 2017. Fairness-aware machine learning: A perspective. arXiv preprint arXiv:1708.00754.

    Google Scholar

  • Cite this article

    William Blanzeisky, Pádraig Cunningham. 2022. Using Pareto simulated annealing to address algorithmic bias in machine learning. The Knowledge Engineering Review 37(1), doi: 10.1017/S0269888922000029
    William Blanzeisky, Pádraig Cunningham. 2022. Using Pareto simulated annealing to address algorithmic bias in machine learning. The Knowledge Engineering Review 37(1), doi: 10.1017/S0269888922000029
    shu
Download PDF

Article Metrics

Article views(63) PDF downloads(48)

Other Articles By Authors

RESEARCH ARTICLE   Open Access    

Using Pareto simulated annealing to address algorithmic bias in machine learning

  • Received: 31 August 2021
  • Revised: 24 March 2022
  • Accepted: 25 March 2022
  • Published online: 04 May 2022
The Knowledge Engineering Review  37,  2022, 37(1)  |  Cite this article

Abstract: Abstract: Algorithmic bias arises in machine learning when models that may have reasonable overall accuracy are biased in favor of ‘good’ outcomes for one side of a sensitive category, for example gender or race. The bias will manifest as an underestimation of good outcomes for the under-represented minority. In a sense, we should not be surprised that a model might be biased when it has not been ‘asked’ not to be; reasonable accuracy can be achieved by ignoring the under-represented minority. A common strategy to address this issue is to include fairness as a component in the learning objective. In this paper, we consider including fairness as an additional criterion in model training and propose a multi-objective optimization strategy using Pareto Simulated Annealing that optimizes for both accuracy and underestimation bias. Our experiments show that this strategy can identify families of models with members representing different accuracy/fairness tradeoffs. We demonstrate the effectiveness of this strategy on two synthetic and two real-world datasets.

    HTML

    Acknowledgments

    • The authors declare none

    Acknowledgments

    • This work was funded by Science Foundation Ireland through the SFI Centre for Research Training in Machine Learning (Grant No.18/CRT/6183) with support from Microsoft Ireland.

    • https://github.com/williamblanzeisky/ParetoSimulatedAnnealing

    • https://scikit-learn.org/

    Rights and permissions
    • This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
References (27)
  • About this article
    Cite this article
    William Blanzeisky, Pádraig Cunningham. 2022. Using Pareto simulated annealing to address algorithmic bias in machine learning. The Knowledge Engineering Review 37(1), doi: 10.1017/S0269888922000029
    William Blanzeisky, Pádraig Cunningham. 2022. Using Pareto simulated annealing to address algorithmic bias in machine learning. The Knowledge Engineering Review 37(1), doi: 10.1017/S0269888922000029
    shu

    • Catalog

      /

      DownLoad:  Full-Size Img  PowerPoint
      Return
      Return