Stability and multigroup fairness in ranking with uncertain predictions
Article No.: 424, Pages 10661 - 10686
Abstract
Rankings are ubiquitous across many applications, from search engines to hiring committees. In practice, many rankings are derived from the output of predictors. However, when predictors trained for classification tasks have intrinsic uncertainty, it is not obvious how this uncertainty should be represented in the derived rankings. Our work considers ranking functions: maps from individual predictions for a classification task to distributions over rankings. We focus on two aspects of ranking functions: stability to perturbations in predictions and fairness towards both individuals and subgroups. Not only is stability an important requirement for its own sake, but--as we show-- it composes harmoniously with individual fairness in the sense of Dwork et al. (2012). While deterministic ranking functions cannot be stable aside from trivial scenarios, we show that the recently proposed uncertainty aware (UA) ranking functions of Singh et al. (2021) are stable. Our main result is that UA rankings also achieve group fairness through successful composition with multiaccurate or multicalibrated predictors. Our work demonstrates that UA rankings naturally interpolate between group and individual level fairness guarantees, while simultaneously satisfying stability guarantees important whenever machine-learned predictions are used.
References
[1]
Angelopoulos, A. N. and Bates, S. A gentle introduction to conformal prediction and distribution-free uncertainty quantification. Preprint, 2021.
[2]
Asudeh, A., Jagadish, H. V., Miklau, G., and Stoyanovich, J. On obtaining stable rankings. In Proc. 44th Intl. Conf. on Very Large Data Bases, volume 12, pp. 237-250, 2018.
[3]
Awasthi, P., Kleindessner, M., and Morgenstern, J. Equalized odds postprocessing under imperfect group information. In Proc. 23rd Intl. Conf. on Artificial Intelligence and Statistics, pp. 1770-1780. PMLR, 2020.
[4]
Bastani, O., Gupta, V., Jung, C., Noarov, G., Ramalingam, R., and Roth, A. Practical adversarial multivalid conformal prediction. In Proc. 36th Advances in Neural Information Processing Systems, 2022.
[5]
Birkhoff, G. Tres observaciones sobre el algebra lineal. Univ. Nac. Tucuman, Ser. A, 5:147-154, 1946.
[6]
Bradley, R. A. and Terry, M. E. Rank analysis of incomplete block designs: I. the method of paired comparisons. Biometrika, 39(3/4):324-345, 1952.
[7]
Bruch, S., Han, S., Bendersky, M., and Najork, M. A stochastic treatment of learning to rank scoring functions. In Proc. 13th ACM Intl. Conf. on Web Search and Data Mining, pp. 61-69, 2020.
[8]
Busa-Fekete, R., Kégl, B., Éltetö, T., and Szarvas, G. Ranking by calibrated adaboost. In Proceedings of the Learning to Rank Challenge, pp. 37-48. PMLR, 2011.
[9]
Cao, Z., Qin, T., Liu, T.-Y., Tsai, M.-F., and Li, H. Learning to rank: from pairwise approach to listwise approach. In Proc. 24th Intl. Conf. on Machine Learning, pp. 129-136, 2007.
[10]
Caton, S. and Haas, C. Fairness in machine learning: A survey. ACM Computing Surveys, 2020.
[11]
Cohen, D., Mitra, B., Lesota, O., Rekabsaz, N., and Eickhoff, C. Not all relevance scores are equal: Efficient uncertainty and calibration modeling for deep retrieval models. In Proc. 44th Intl. Conf. on Research and Development in Information Retrieval (SIGIR), pp. 654-664, 2021.
[12]
Cooper, A. F., Lee, K., Barocas, S., De Sa, C., Sen, S., and Zhang, B. Is my prediction arbitrary? Measuring self-consistency in fair classification. In Proc. 38th AAAI Conf. on Artificial Intelligence, 2024.
[13]
Devic, S., Kempe, D., Sharan, V., and Korolova, A. Fairness in matching under uncertainty. In Proc. 40th Intl. Conf. on Machine Learning, volume 202, pp. 7775-7794, 2023.
[14]
DiCiccio, C., Hsu, B., Yu, Y., Nandy, P., and Basu, K. Detection and mitigation of algorithmic bias via predictive parity. In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, pp. 1801-1816, 2023.
[15]
Ding, F., Hardt, M., Miller, J., and Schmidt, L. Retiring adult: New datasets for fair machine learning. In Proc. 35th Advances in Neural Information Processing Systems, 2021.
[16]
Dwork, C., Hardt, M., Pitassi, T., Reingold, O., and Zemel, R. Fairness through awareness. In Proc. 3rd Innovations in Theoretical Computer Science, pp. 214-226, 2012.
[17]
Dwork, C., Kim, M. P., Reingold, O., Rothblum, G. N., and Yona, G. Learning from outcomes: Evidence-based rankings. In Proc. 60th IEEE Symp. on Foundations of Computer Science, pp. 106-125. IEEE, 2019.
[18]
Ganesh, P., Chang, H., Strobel, M., and Shokri, R. On the impact of machine learning randomness on group fairness. In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, pp. 1789-1800, 2023.
[19]
García-Soriano, D. and Bonchi, F. Maxmin-fair ranking: individual fairness under group-fairness constraints. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, pp. 436-446, 2021.
[20]
Geyik, S. C., Ambler, S., and Kenthapadi, K. Fairnessaware ranking in search & recommendation systems with application to LinkedIn talent search. In Proc. 25th Intl. Conf. on Knowledge Discovery and Data Mining, pp. 2221-2231, 2019.
[21]
Goodfellow, I. J., Shlens, J., and Szegedy, C. Explaining and harnessing adversarial examples. In Bengio, Y. and Le-Cun, Y. (eds.), 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, 2015. URL http://arxiv.org/abs/1412.6572. 5
[22]
Google. Ads help: About ad rank, 2023. URL https://support.google.com/google-ads/answer/1722122?hl=en. Accessed: 2023-09-28.
[23]
Gopalan, P., Kim, M. P., Singhal, M. A., and Zhao, S. Low-degree multicalibration. In Proc. 35th Conference on Learning Theory, pp. 3193-3234. PMLR, 2022.
[24]
Gorantla, S., Mehrotra, A., Deshpande, A., and Louis, A. Sampling individually-fair rankings that are always group fair. In Rossi, F., Das, S., Davis, J., Firth-Butterfield, K., and John, A. (eds.), Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, AIES 2023, pp. 205-216. ACM, 2023.
[25]
Guiver, J. and Snelson, E. Learning to rank with softrank and gaussian processes. In Proc. 31st Intl. Conf. on Research and Development in Information Retrieval (SIGIR), pp. 259-266, 2008.
[26]
Guo, C., Pleiss, G., Sun, Y., and Weinberger, K. Q. On calibration of modern neural networks. In Proc. 34th Intl. Conf. on Machine Learning, pp. 1321-1330. PMLR, 2017.
[27]
Guo, R., Ton, J.-F., and Liu, Y. Fair learning to rank with distribution-free risk control. Preprint, 2023.
[28]
Gupta, C. and Ramdas, A. Top-label calibration and multiclass-to-binary reductions. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net, 2022. URL https://openreview.net/forum?id=WqoBaaPHS-.
[29]
Haghtalab, N., Jordan, M. I., and Zhao, E. A unifying perspective on multi-calibration: Unleashing game dynamics for multi-objective learning. Proc. 37th Advances in Neural Information Processing Systems, 36, 2023.
[30]
Hardt, M., Price, E., and Srebro, N. Equality of opportunity in supervised learning. Proc. 30th Advances in Neural Information Processing Systems, 29, 2016.
[31]
Harper, F. M. and Konstan, J. A. The movielens datasets: History and context. ACM Trans. Interact. Intell. Syst., 2015.
[32]
Hébert-Johnson, U., Kim, M., Reingold, O., and Rothblum, G. Multicalibration: Calibration for the (computationally-identifiable) masses. In Proc. 35th Intl. Conf. on Machine Learning, pp. 1939-1948. PMLR, 2018.
[33]
Heuss, M., Cohen, D., Mansoury, M., Rijke, M. d., and Eickhoff, C. Predictive uncertainty-based bias mitigation in ranking. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, pp. 762-772, 2023.
[34]
Huang, L. and Vishnoi, N. Stable and fair classification. In Proc. 36th Intl. Conf. on Machine Learning, 2019.
[35]
Järvelin, K. and Kekäläinen, J. Cumulated gain-based evaluation of ir techniques. ACM Transactions on Information Systems (TOIS), 20(4):422-446, 2002.
[36]
Jung, C., Noarov, G., Ramalingam, R., and Roth, A. Batch multivalid conformal prediction. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023, 2023.
[37]
Kim, M. P., Ghorbani, A., and Zou, J. Multiaccuracy: Blackbox post-processing for fairness in classification. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society, pp. 247-254, 2019.
[38]
Korevaar, H., McConnell, C., Tong, E., Brinkman, E., Shine, A., Abbas, M., Metevier, B., Corbett-Davies, S., and El-Arini, K. Matched pair calibration for ranking fairness. arXiv preprint arXiv:2306.03775, 2023.
[39]
Kweon, W., Kang, S., and Yu, H. Obtaining calibrated probabilities with personalized ranking models. In Proc. 36th AAAI Conf. on Artificial Intelligence, volume 36, pp. 4083-4091, 2022.
[40]
Luce, R. D. Individual choice behavior. Courier Corporation, 1959.
[41]
Martins, M. V., Tolledo, D., Machado, J., Baptista, L. M., and Realinho, V. Early prediction of student's performance in higher education: A case study. In Trends and Applications in Information Systems and Technologies: Volume 1 9, pp. 166-175. Springer, 2021
[42]
Mehrotra, A. and Celis, L. E. Mitigating bias in set selection with noisy protected attributes. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp. 237-248, 2021.
[43]
Mehrotra, A. and Vishnoi, N. Fair ranking with noisy protected attributes. Proc. 36th Advances in Neural Information Processing Systems, 35:31711-31725, 2022.
[44]
Menon, A. K., Jiang, X. J., Vembu, S., Elkan, C., and Ohno-Machado, L. Predicting accurate probabilities with a ranking loss. In Proc. 29th Intl. Conf. on Machine Learning, volume 2012, pp. 703, 2012.
[45]
Meta. Our approach to facebook feed ranking, 2023. URL https://transparency.fb.com/features/ranking-and-content/. Accessed: 2023-09-28.
[46]
Minderer, M., Djolonga, J., Romijnders, R., Hubis, F., Zhai, X., Houlsby, N., Tran, D., and Lucic, M. Revisiting the calibration of modern neural networks. Proc. 35th Advances in Neural Information Processing Systems, 34: 15682-15694, 2021.
[47]
Narasimhan, H., Cotter, A., Gupta, M., and Wang, S. Pairwise fairness for ranking and regression. In Proc. 34th AAAI Conf. on Artificial Intelligence, volume 34, pp. 5248-5255, 2020.
[48]
Nettleton, D. F., Orriols-Puig, A., and Fornells, A. A study of the effect of different types of noise on the precision of supervised learning techniques. Artificial intelligence review, 33:275-306, 2010.
[49]
Oh, S., Ustun, B., McAuley, J., and Kumar, S. Rank list sensitivity of recommender systems to interaction perturbations. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pp. 1584-1594, 2022.
[50]
Oh, S., Ustun, B., McAuley, J., and Kumar, S. Finest: Stabilizing recommendations by rank-preserving fine-tuning. arXiv preprint arXiv:2402.03481, 2024.
[51]
Penha, G. and Hauff, C. On the calibration and uncertainty of neural learning to rank models for conversational search. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pp. 160-170, 2021.
[52]
Pitoura, E., Stefanidis, K., and Koutrika, G. Fairness in rankings and recommendations: an overview. In Proc. 48th Intl. Conf. on Very Large Data Bases, 2022.
[53]
Plackett, R. L. The analysis of permutations. Journal of the Royal Statistical Society Series C: Applied Statistics, 24 (2):193-202, 1975.
[54]
Rastogi, R. and Joachims, T. Fair ranking under disparate uncertainty. Preprint, 2023.
[55]
Robertson, S. E. The probability ranking principle in ir. Journal of documentation, 33(4):294-304, 1977.
[56]
Shabat, E., Cohen, L., and Mansour, Y. Sample complexity of uniform convergence for multicalibration. Proc. 34th Advances in Neural Information Processing Systems, 33: 13331-13340, 2020.
[57]
Shalev-Shwartz, S. and Ben-David, S. Understanding machine learning: From theory to algorithms. Cambridge university press, 2014.
[58]
Shen, Z., Wang, Z., Zhu, X., Fain, B., and Munagala, K. Fairness in the assignment problem with uncertain priorities. In Proc. 22nd Intl. Conf. on Autonomous Agents and Multiagent Systems, pp. 188-196, 2023.
[59]
Singh, A. and Joachims, T. Fairness of exposure in rankings. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, pp. 2219-2228, 2018.
[60]
Singh, A. and Joachims, T. Policy learning for fairness in ranking. Proc. 33rd Advances in Neural Information Processing Systems, 32, 2019.
[61]
Singh, A., Kempe, D., and Joachims, T. Fairness in ranking under uncertainty. In Proc. 35th Advances in Neural Information Processing Systems, pp. 11896-11908, 2021.
[62]
Soliman, M. A. and Ilyas, I. F. Ranking with uncertain scores. In 2009 IEEE 25th international conference on data engineering, pp. 317-328. IEEE, 2009.
[63]
Tahir, A., Cheng, L., and Liu, H. Fairness through aleatoric uncertainty. Preprint, 2023.
[64]
Tang, B., Koçyiğit, C,., Rice, E., and Vayanos, P. Learning optimal and fair policies for online allocation of scarce societal resources from data collected in deployment. arXiv preprint arXiv:2311.13765, 2023.
[65]
Taylor, M., Guiver, J., Robertson, S., and Minka, T. Softrank: optimizing non-smooth rank metrics. In Proc. 1st ACM Intl. Conf. on Web Search and Data Mining, pp. 77-86, 2008.
[66]
Thurstone, L. L. A law of comparative judgment. Psychological review, 101(2):266, 1994.
[67]
TurboHire. Unleashing the power of ai & automation to effortlessly discover the best talent, 2023. URL https://turbohire.co/features/talent-screening/#candidate-scoring. Accessed: 2023-10-09.
[68]
Wang, C.-J. and Chen, H.-H. Learning to predict the costper-click for your ad words. In Proceedings of the 21st ACM international conference on Information and knowledge management, pp. 2291-2294, 2012.
[69]
Wilde, N., Biyik, E., Sadigh, D., and Smith, S. L. Learning reward functions from scale feedback. Proceedings of the 5th Conference on Robot Learning (CoRL), 2021.
[70]
Xu, J. and Li, H. Adarank: a boosting algorithm for information retrieval. In Proc. 30th Intl. Conf. on Research and Development in Information Retrieval (SIGIR), pp. 391-398, 2007.
[71]
Yan, L., Qin, Z., Wang, X., Bendersky, M., and Najork, M. Scale calibration of deep ranking models. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 4300-4309, 2022.
[72]
Yang, T., Luo, C., Lu, H., Gupta, P., Yin, B., and Ai, Q. Can clicks be both labels and features? unbiased behavior feature collection and uncertainty-aware learning to rank. In Proc. 45th Intl. Conf. on Research and Development in Information Retrieval (SIGIR), pp. 6-17, 2022.
[73]
Yang, T., Xu, Z., Wang, Z., Tran, A., and Ai, Q. Marginal-certainty-aware fair ranking algorithm. In Proc. 16th ACM Intl. Conf. on Web Search and Data Mining, pp. 24-32, 2023.
[74]
Zehlike, M., Yang, K., and Stoyanovich, J. Fairness in ranking: A survey. Preprint, 2021. URL https://arxiv.org/abs/2103.14000.
[75]
Zhu, B., Jordan, M., and Jiao, J. Principled reinforcement learning with human feedback from pairwise or k-wise comparisons. In Proc. 40th Intl. Conf. on Machine Learning, pp. 43037-43067. PMLR, 2023.
[76]
Zou, A., Wang, Z., Kolter, J. Z., and Fredrikson, M. Universal and transferable adversarial attacks on aligned language models. Preprint, 2023.
Index Terms
- Stability and multigroup fairness in ranking with uncertain predictions
Index terms have been assigned to the content through auto-classification.
Recommendations
A less conservative robust stability criteria for uncertain neutral systems with mixed delays
This paper is concerned with the problem of the delay-dependent robust stability of neutral systems with mixed delays and time-varying structured uncertainties. By considering the cross-terms with additional design parameters, a complete form of ...
Stochastic Stability for Uncertain Neutral Markovian Jump Systems with Nonlinear Perturbations
The delay-range-dependent stochastic stability for uncertain neutral Markovian jump systems with interval time-varying delays and nonlinear perturbations is investigated. The perturbations under consideration are time-varying and norm-bounded. By delay ...
Novel robust stability criteria for uncertain systems with time-varying delay
This paper is concerned with the delay-dependent stability and robust stability criteria for linear systems with time-varying delay and norm-bounded uncertainties. Through constructing a general form of Lyapunov-Krasovskii functional, and using integral ...
Comments
Information & Contributors
Information
Published In
July 2024
63010 pages
Copyright © 2024.
Publisher
JMLR.org
Publication History
Published: 21 July 2024
Qualifiers
- Research-article
- Research
- Refereed limited
Acceptance Rates
Overall Acceptance Rate 140 of 548 submissions, 26%
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 0Total Downloads
- Downloads (Last 12 months)0
- Downloads (Last 6 weeks)0
Reflects downloads up to 26 Jan 2025