{"type":"ScholarlyArticle","title":"Individual Fairness Guarantees for Neural Networks","authors":[{"name":"Elias Benussi","slug":"elias-benussi","given":"Elias","family":"Benussi","affiliation":"University of Oxford"},{"name":"Andrea Patane'","slug":"andrea-patane","given":"Andrea","family":"Patane","affiliation":"University of Oxford"},{"name":"Matthew Wicker","slug":"matthew-wicker","given":"Matthew","family":"Wicker","affiliation":"University of Oxford"},{"name":"Luca Laurenti","slug":"luca-laurenti","given":"Luca","family":"Laurenti","affiliation":"TU Delft"},{"name":"Marta Kwiatkowska","slug":"marta-kwiatkowska","given":"Marta","family":"Kwiatkowska","affiliation":"University of Oxford"}],"abstract":"We consider the problem of certifying the individual fairness (IF) of feed-forward neural networks (NNs). \r\nIn particular, we work with the epsilon-delta-IF formulation, which, given a NN and a similarity metric learnt from data, requires that the output difference between any pair of epsilon-similar individuals is bounded by a maximum decision tolerance delta >= 0. \r\nWorking with a range of metrics, including the Mahalanobis distance, we propose a method to overapproximate the resulting optimisation problem using piecewise-linear functions to lower and upper bound the NN's non-linearities globally over the input space.\r\nWe encode this computation as the solution of a Mixed-Integer Linear Programming problem and demonstrate that it can be used to compute IF guarantees on four datasets widely used for fairness benchmarking.\r\nWe show how this formulation can be used to encourage models' fairness at training time by modifying the NN loss, and empirically confirm our approach yields NNs that are orders of magnitude fairer than state-of-the-art methods.","doi":"10.24963/ijcai.2022/92","url":"https://ijcai.org/proceedings/2022/92","pdf":"https://ijcai.org/proceedings/2022/0092.pdf","markdown":"https://ijcai.org/proceedings/2022/92.md","bibtex":"https://ijcai.org/proceedings/2022/bibtex/92","volume":{"slug":"2022","title":"Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2022","url":"https://ijcai.org/proceedings/2022/","year":2022,"location":"Vienna","dates":"23-29 July 2022","editors":"Luc De Raedt","isbn":"978-1-956792-00-3","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2022"},"section":"AI Ethics, Trust, Fairness","track":"Main Track","number":92,"pages":{"first":651,"last":658},"keywords":["AI Ethics, Trust, Fairness: Fairness & Diversity"],"videos":[{"label":"Video #1","url":"https://slideslive.com/38987062","length_sec":86,"provider":"slideslive"},{"label":"Video #2","url":"https://slideslive.com/38985205","length_sec":641,"provider":"slideslive"}],"references":[{"position":1,"raw":"[Albarghouthi et al., 2017] Aws Albarghouthi, Loris D'Antoni, Samuel Drews, and Aditya V Nori. Fairsquare: probabilistic verification of program fairness. Pro- ceedings of the ACM on Programming Languages, 1(OOPSLA):1-30, 2017.","title":"Fairsquare: probabilistic verification of program fairness","authors":"Albarghouthi","year":"2017","venue":"Proceedings of the ACM on Programming Languages","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Angwin et al., 2016] Julia Angwin, Jeff Larson, Surya Mattu, and Lauren Kirchner. Machine bias. In Ethics of Data and Analytics, pages 254-264.","title":"Machine bias","authors":"Julia Angwin, Jeff Larson, Surya Mattu, Lauren Kirchner","year":null,"venue":"Ethics of Data and Analytics","doi":null,"url":null,"internal_url":null},{"position":3,"raw":"[Barocas and Selbst, 2016] Solon Barocas and Andrew D Selbst. Big data's disparate impact. Calif. L. Rev., 104:671-733, 2016.","title":"Big data's disparate impact","authors":"Selbst Barocas, Solon Barocas, Andrew D Selbst","year":"2016","venue":"Calif. L. Rev","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Bastani et al., 2019] Osbert Bastani, Xin Zhang, and Ar- mando Solar-Lezama. Probabilistic verification of fairness properties via concentration. Proceedings of the ACM on Programming Languages, 3(OOPSLA):1-27, 2019.","title":"Probabilistic verification of fairness properties via concentration","authors":"Osbert Bastani, Xin Zhang, Armando Solar-Lezama","year":"2019","venue":"Proceedings of the ACM on Programming Languages","doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Bolukbasi et al., 2016] Tolga Bolukbasi, Kai-Wei Chang, James Y Zou, Venkatesh Saligrama, and Adam T Kalai. Man is to computer programmer as woman is to home- maker? debiasing word embeddings. NeurIPS, 29:4349- 4357, 2016.","title":"Man is to computer programmer as woman is to homemaker? debiasing word embeddings","authors":"Tolga Bolukbasi, Kai-Wei Chang, James Y Zou, Venkatesh Saligrama, Adam T Kalai","year":"2016","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Boopathy et al., 2019] Akhilan Boopathy, Tsui-Wei Weng, Pin-Yu Chen, Sijia Liu, and Luca Daniel. Cnn-cert: An efficient framework for certifying robustness of convolu- tional neural networks. In AAAI, volume 33, pages 3240- 3247, 2019.","title":"Cnn-cert: An efficient framework for certifying robustness of convolutional neural networks","authors":"Boopathy","year":"2019","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[Dantzig, 2016] George Dantzig. Linear programming and extensions. Princeton university press, 2016.","title":"Linear programming and extensions","authors":"George Dantzig","year":"2016","venue":null,"doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Del Pia and Weismantel, 2012] Alberto Del Pia and Robert Weismantel. On convergence in mixed integer pro- gramming. Mathematical programming, 135(1):397-412, 2012.","title":"Alberto Del Pia and Robert Weismantel. On convergence in mixed integer programming","authors":"Del Pia, Weismantel","year":"2012","venue":"Mathematical programming","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Dwork et al., 2012] Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Richard Zemel. Fairness through awareness. In Proceedings of the 3rd innovations in theoretical computer science conference, pages 214- 226, 2012.","title":"Fairness through awareness","authors":"Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, Richard Zemel","year":"2012","venue":"Proceedings of the 3rd innovations in theoretical computer science conference","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Ehlers, 2017] Ruediger Ehlers. Formal verification of piece- wise linear feed-forward neural networks. In International Symposium on Automated Technology for Verification and Analysis, pages 269-286. Springer, 2017.","title":"Formal verification of piecewise linear feed-forward neural networks","authors":"Ruediger Ehlers","year":"2017","venue":"International Symposium on Automated Technology for Verification and Analysis","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Goodfellow et al., 2015] Ian J. Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. In ICLR, 2015.","title":"Explaining and harnessing adversarial examples","authors":"Ian J Goodfellow, Jonathon Shlens, Christian Szegedy","year":"2015","venue":"ICLR","doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Hardt et al., 2016] Moritz Hardt, Eric Price, and Nati Sre- bro. Equality of opportunity in supervised learning. NeurIPS, 29:3315-3323, 2016.","title":"Equality of opportunity in supervised learning","authors":"Moritz Hardt, Eric Price, Nati Srebro","year":"2016","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Ilvento, 2020] Christina Ilvento. Metric learning for indi- vidual fairness. In 1st Symposium on Foundations of Re- sponsible Computing, 2020.","title":"Metric learning for individual fairness","authors":"Christina Ilvento","year":"2020","venue":"1st Symposium on Foundations of Responsible Computing","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[John et al., 2020] Philips George John, Deepak Vijay- keerthy, and Diptikalyan Saha. Verifying individual fair- ness in machine learning models. In UAI, pages 749-758.","title":"Philips George John, Deepak Vijaykeerthy, and Diptikalyan Saha. Verifying individual fairness in machine learning models","authors":"John","year":"2020","venue":"UAI","doi":null,"url":null,"internal_url":null},{"position":15,"raw":"PMLR, 2020.","title":null,"authors":"Pmlr","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[Katz et al., 2017] Guy Katz, Clark Barrett, David L Dill, Kyle Julian, and Mykel J Kochenderfer. Reluplex: An efficient smt solver for verifying deep neural networks. In International conference on computer aided verification, pages 97-117. Springer, 2017.","title":"Reluplex: An efficient smt solver for verifying deep neural networks","authors":"Katz","year":"2017","venue":"International conference on computer aided verification","doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Kirk et al., 2021] Hannah Rose Kirk, Filippo Volpin, Haider Iqbal, Elias Benussi, Frederic Dreyer, Aleksandar Shtedritski, Yuki Asano, et al. Bias out-of-the-box: An empirical analysis of intersectional occupational biases in popular generative language models. NeurIPS, 34:2611- 2624, 2021.","title":"Bias out-of-the-box: An empirical analysis of intersectional occupational biases in popular generative language models","authors":"Rose Hannah, Filippo Kirk, Haider Volpin, Elias Iqbal, Frederic Benussi, Aleksandar Dreyer, Yuki Shtedritski, Asano","year":"2021","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Leino et al., 2021] Klas Leino, Zifan Wang, and Matt Fredrikson. Globally-robust neural networks. In Inter- national Conference on Machine Learning, pages 6212- 6222. PMLR, 2021.","title":"Globally-robust neural networks","authors":"Klas Leino, Zifan Wang, Matt Fredrikson","year":"2021","venue":"International Conference on Machine Learning","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Madry et al., 2018] Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks. In ICLR, 2018.","title":"Towards deep learning models resistant to adversarial attacks","authors":"Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, Adrian Vladu","year":"2018","venue":"ICLR","doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[McNamara et al., 2017] Daniel McNamara, Cheng Soon Ong, and Robert C. Williamson. Provably fair represen- tations. CoRR, abs/1710.04394, 2017.","title":"Provably fair representations","authors":"Daniel Mcnamara, Cheng, Soon Ong, Robert C Williamson","year":"2017","venue":null,"doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Mukherjee et al., 2020] Debarghya Mukherjee, Mikhail Yurochkin, Moulinath Banerjee, and Yuekai Sun. Two simple ways to learn individual fairness metrics from data. In International Conference on Machine Learning, pages 7097-7107. PMLR, 2020.","title":"Two simple ways to learn individual fairness metrics from data","authors":"Mukherjee","year":"2020","venue":"International Conference on Machine Learning","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Ruoss et al., 2020] Anian Ruoss, Mislav Balunovic, Marc Fischer, and Martin Vechev. Learning certified individu- ally fair representations. NeurIPS, 33:7584-7596, 2020.","title":"Learning certified individually fair representations","authors":"Anian Ruoss, Mislav Balunovic, Marc Fischer, Martin Vechev","year":"2020","venue":"NeurIPS","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Tjeng et al., 2018] Vincent Tjeng, Kai Y Xiao, and Russ Tedrake. Evaluating robustness of neural networks with mixed integer programming. In ICLR, 2018.","title":"Evaluating robustness of neural networks with mixed integer programming","authors":"Vincent Tjeng, Kai Y Xiao, Russ Tedrake","year":"2018","venue":"ICLR","doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Urban et al., 2020] Caterina Urban, Maria Christakis, Valentin Wüstholz, and Fuyuan Zhang. Perfectly par- allel fairness certification of neural networks. ACM on Programming Languages, 4(OOPSLA):1-30, 2020.","title":"Perfectly parallel fairness certification of neural networks","authors":"Caterina Urban, Maria Christakis, Valentin Wüstholz, Fuyuan Zhang","year":"2020","venue":"ACM on Programming Languages","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Wicker et al., 2020] Matthew Wicker, Luca Laurenti, An- drea Patane, and Marta Kwiatkowska. Probabilistic safety for bayesian neural networks. In UAI, pages 1198-1207.","title":"Probabilistic safety for bayesian neural networks","authors":"Matthew Wicker, Luca Laurenti, Andrea Patane, Marta Kwiatkowska","year":null,"venue":"UAI","doi":null,"url":null,"internal_url":null},{"position":26,"raw":"PMLR, 2020.","title":null,"authors":"Pmlr","year":"2020","venue":null,"doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[Wicker et al., 2021] Matthew Wicker, Luca Laurenti, An- drea Patane, Zhuotong Chen, Zheng Zhang, and Marta Kwiatkowska. Bayesian inference with certifiable adver- sarial robustness. In AISTATS, pages 2431-2439. PMLR, 2021.","title":"Bayesian inference with certifiable adversarial robustness","authors":"Matthew Wicker, Luca Laurenti, Andrea Patane, Zhuotong Chen, Zheng Zhang, Marta Kwiatkowska","year":"2021","venue":"AISTATS","doi":null,"url":null,"internal_url":null},{"position":28,"raw":"[Yeom and Fredrikson, 2021] Samuel Yeom and Matt Fredrikson. Individual fairness revisited: transferring techniques from adversarial robustness. In IJCAI, pages 437-443, 2021.","title":"Individual fairness revisited: transferring techniques from adversarial robustness","authors":"Fredrikson Yeom, Samuel Yeom, Matt Fredrikson","year":"2021","venue":"IJCAI","doi":null,"url":null,"internal_url":"/proceedings/2020/61"},{"position":29,"raw":"[Yurochkin et al., 2020] Mikhail Yurochkin, Amanda Bower, and Yuekai Sun. Training individually fair ml models with sensitive subspace robustness. In ICLR, pages 1-18, 2020.","title":"Training individually fair ml models with sensitive subspace robustness","authors":"Mikhail Yurochkin, Amanda Bower, Yuekai Sun","year":"2020","venue":"ICLR","doi":null,"url":null,"internal_url":null}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}