Aniket Kesari
ABSTRACT
Both law and computer science are concerned with developing frameworks for protecting privacy and ensuring fairness. Both fields often consider these two values separately and develop legal doctrines and machine learning metrics in isolation from one another. Yet, privacy and fairness values can conflict, especially when considered alongside the accuracy of an algorithm. The computer science literature often treats this problem as an "impossibility theorem" - we can have privacy or fairness but not both. Legal doctrine is similarly constrained by a focus on the inputs to a decision - did the decisionmaker intend to use information about protected attributes. Despite these challenges, there is a way forward. The law has integrated economic frameworks to consider tradeoffs in other domains, and a similar approach can clarify policymakers' thinking around balancing accuracy, privacy, and fairnesss. This piece illustrates this idea by using a law & economics lens to formalize the notion of a Privacy-Fairness-Accuracy frontier, and demonstrating this framework on a consumer lending dataset. An open-source Python software library and GUI will be made available.
Supplemental Material
- [n.d.].Google Scholar
- 1968. Truth in Lending Act. 15 U.S.C. ch. 41 § 1601.Google Scholar
- 1970. Fair Credit Reporting Act. 12 U.S.C. ch. 16 §§ 1830--1831; 15 U.S.C. ch. 41 § 1681 et seq.Google Scholar
- 1974. Equal Credit Opportunity Act. 15 U.S.C. § 1691 et seq.Google Scholar
- Sushant Agarwal. 2021. Trade-Offs between Fairness and Privacy in Machine Learning. IJCAI 2021 Workshop on AI for Social Good (2021).Google Scholar
- Geoffrey M.B. Tootell Alicia H. Munnell, Lynn E. Browne, James McEneaney. 1992. Mortgage Lending in Boston: Interpreting HMDA Data. Federal Reserve of Boston Working Paper Series October (1992).Google Scholar
- Eugene Bagdasaryan, Omid Poursaeed, and Vitaly Shmatikov. 2019. Differential privacy has disparate impact on model accuracy. Advances in Neural Information Processing Systems 32, NeurIPS (2019). arXiv:1905.12101Google Scholar
- S. Barocas, M. Hardt, and A. Narayanan. 2019. Limitations and Opportunities Solon Barocas , Moritz Hardt , Arvind Narayanan. Fairness and Machine Learning: Limitation and Oppotunities (2019). https://fairmlbook.orgGoogle Scholar
- Andrew D. Barocas, Solon; Selbst. 2016. Big Data's Disparate Impact. California Law Review 104, 671 (2016), 671--732.Google Scholar
- Hongyan Chang and Reza Shokri. 2021. On the privacy risks of algorithmic fairness. Proceedings - 2021 IEEE European Symposium on Security and Privacy, Euro S and P 2021 (2021), 292--303. https://doi.org/10.1109/EuroSP51992.2021. 00028 arXiv:2011.03731Google ScholarCross Ref
- Robert Cooter and Thomas Ulen. 2016. Law and Economics, 6th Edition. Berkeley Law Books.Google Scholar
- Adam Smith Cynthia Dwork, Frank McSherry, Kobbi Nissim. 2006. Calibrating Noise to Sensitivity in Private Data Analysis. In Theory of Cryptography. 265--284. https://doi.org/10.1109/IECON.2017.8216701Google ScholarDigital Library
- John J. Donohue. 2005. The Law and Economics of Antidiscrimination Law. Handbook of Law and Economics July (2005), 1--84.Google ScholarCross Ref
- Cynthia Dwork and Aaron Roth. 2013. The algorithmic foundations of differential privacy. , 211--487 pages. https://doi.org/10.1561/0400000042Google ScholarDigital Library
- Michael D. Ekstrand, Rezvan Joshaghani, and Hoda Mehrpouyan. 2018. Privacy for All: Ensuring Fair and Equitable Privacy Protections. Proceedings of Machine Learning Research 81 (2018), 35--47. https://proceedings.mlr.press/v81/ ekstrand18a.htmlGoogle Scholar
- Talia B. Gillis. 2021. The Input Fallacy. Minnesota Law Review (2021).Google Scholar
- Naoise Holohan, Stefano Braghin, Pól Mac Aonghusa, and Killian Levacher. 2019. Diffprivlib: the IBM differential privacy library. ArXiv e-prints 1907.02444 [cs.CR] (July 2019).Google Scholar
- Naoise Holohan, Stefano Braghin, Pól Mac Aonghusa, and Killian Levacher. 2019. Diffprivlib: The IBM Differential Privacy Library. (2019), 1--5. arXiv:1907.02444 http://arxiv.org/abs/1907.02444Google Scholar
- Richard M. Hynes and Eric A. Posner. 2002. Law and Economics of Consumer Finance. American Law and Economics Review 4, 1 (2002), 168--207. https: //doi.org/10.2139/ssrn.261109Google ScholarCross Ref
- Matthew Jagielski, Michael Kearns, Jieming Mao, Alina Oprea, Aaron Roth, Saeed Sharifi-Malvajerdi, and Jonathan Ullman. 2019. Differentially private fair learning. 36th International Conference on Machine Learning, ICML 2019 2019-June (2019), 5306--5314. arXiv:1812.02696Google Scholar
- Zachary Liscow. 2018. Is Efficiency Biased ? University of Chicago Law Review 85 (2018).Google Scholar
- Rory Van Loo. 2021. Privacy Pretexts. Cornell Law Review (2021).Google Scholar
- Ashwin Machanavajjhala, Daniel Kifer, John Abowd, Johannes Gehrke, and Lars Vilhuber. 2008. Privacy: Theory meets practice on the map. Proceedings - International Conference on Data Engineering 2003 (2008), 277--286. https: //doi.org/10.1109/ICDE.2008.4497436Google ScholarDigital Library
- Tabish Maniar, Alekhya Akkinepally, and Anantha Sharma. 2021. Differential Privacy for Credit Risk Model. (2021), 1--7. arXiv:2106.15343 http://arxiv.org/ abs/2106.15343Google Scholar
- Cynthia Dwork Deirdre K. Mulligan. 2013. It's Not Privacy, and It's Not Fair. Stanford Law Review Online 66, 2000 (2013), 35. http://www.stanfordlawreview.org/online/privacy-and-big-data/its-notprivacy- and-its-not-fair%5Cnhttp://www.stanfordlawreview.org/sites/default/ files/online/topics/DworkMullliganSLR.pdfGoogle Scholar
- Marlotte Pannekoek and Giacomo Spigler. 2021. Investigating Trade-offs in Utility, Fairness and Differential Privacy in Neural Networks. (2021). arXiv:2102.05975 http://arxiv.org/abs/2102.05975Google Scholar
- Richard L. Revesz. 2018. Regulation and distribution. New York University Law Review 93, 6 (2018), 1489--1578. https://doi.org/10.2139/ssrn.3133192Google Scholar
- Jessica L. Roberts. 2013. Protecting Privacy to Prevent Discrimination. Ssrn 56, 6 (2013). https://doi.org/10.2139/ssrn.2263583Google Scholar
- Pedro Saleiro, Benedict Kuester, Loren Hinkson, Jesse London, Abby Stevens, Ari Anisfeld, Kit T. Rodolfa, and Rayid Ghani. 2018. Aequitas: A Bias and Fairness Audit Toolkit. 2018 (2018). arXiv:1811.05577 http://arxiv.org/abs/1811.05577Google Scholar
- Kailash Karthik Saravanakumar. 2020. The Impossibility Theorem of Machine Fairness -- A Causal Perspective. (2020). arXiv:2007.06024 http://arxiv.org/abs/ 2007.06024Google Scholar
- Lior Jacob Strahilevitz. 2008. Privacy versus antidiscrimination. University of Chicago Law Review 75, 1 (2008), 363--381.Google Scholar
- Rebecca Wexler. 2018. Liberty, and Trade Secrets: Criminal Justice System. Stanford Law Review Life 70, May (2018), 1343--1429. https://review.law.stanford.edu/ wp-content/uploads/sites/3/2018/06/70-Stan.-L.-Rev.-1343.pdfGoogle Scholar
- Alexandra Wood, Micah Altman, Aaron Bembenek, Mark Bun, Marco Gaboardi, James Honaker, Kobbi Nissim, David O'Brien, Thomas Steinke, and Salil Vadhan. 2019. Differential Privacy: A Primer for a Non-Technical Audience. SSRN Electronic Journal (2019). https://doi.org/10.2139/ssrn.3338027Google Scholar
- Depeng Xu. 2021. Achieving Differential Privacy and Fairness in Machine Learning. (2021). https://scholarworks.uark.edu/etd/3960Google Scholar
- Yuli Zheng, Zhenyu Wu, Ye Yuan, Tianlong Chen, and Zhangyang Wang. 2020. PCAL: A Privacy-preserving Intelligent Credit Risk Modeling Framework Based on Adversarial Learning. Vol. 1. Association for Computing Machinery. arXiv:2010.02529 http://arxiv.org/abs/2010.02529Google Scholar
Index Terms
- The Privacy-Fairness-Accuracy Frontier: A Computational Law & Economics Toolkit for Making Algorithmic Tradeoffs
Recommendations
Examining the Gramm–Leach–Bliley Act's opt-out method for protecting consumer data privacy rights on the Internet
This article involves examining the Gramm–Leach–Bliley Act's GLBA opt-out method in concern with protecting consumer data privacy rights online. Its purpose is to show that further regulatory measures are necessary at the federal level in order to ...
RFID system with fairness within the framework of security and privacy
ESAS'05: Proceedings of the Second European conference on Security and Privacy in Ad-Hoc and Sensor NetworksRadio Frequency Identification (RFID) systems are expected to be widely deployed in automated identification and supply-chain applications. Although RFID systems have several advantages, the technology may also create new threats to user privacy. In ...
Digital fraud: a market regulation approach to digital rights/restrictions management
LawTech '07: Proceedings of the Fifth IASTED International Conference on Law and TechnologyDigital Rights/Restrictions Management (hereafter DRM) provokes strong reactions. Some view DRM as a potential threat to free expression. Others view DRM as a technological solution to bargaining failure, obviating some of the need for the "fair use" ...
Comments