Bilel Benbouzid
ABSTRACT
We argue in this article that the integration of fairness into machine learning, or FairML, is a valuable exemplar of the politics of statistics and their ongoing transformations. Classically, statisticians sought to eliminate any trace of politics from their measurement tools. But data scientists who are developing predictive machines for social applications – are inevitably confronted with the problem of fairness. They thus face two difficult and often distinct types of demands: first, for reliable computational techniques, and second, for transparency, given the constructed, politically situated nature of quantification operations. We begin by socially localizing the formation of FairML as a field of research and describing the associated epistemological framework. We then examine how researchers simultaneously think the mathematical and social construction of approaches to machine learning, following controversies around fairness metrics and their status. Thirdly and finally, we show that FairML approaches tend towards a specific form of objectivity, “trained judgement,” which is based on a reasonably partial justification from the designer of the machine – which itself comes to be politically situated as a result.
- J. Kleinberg, J. Ludwig, S. Mullainathan, and C. R. Sunstein, ‘Discrimination in the Age of Algorithms’, Journal of Legal Analysis, vol. 10, pp. 113–174, Dec. 2018, doi: 10.1093/jla/laz001.Google Scholar
Cross Ref
- V. Eubanks, Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. St. Martin's Publishing Group, 2018.Google ScholarDigital Library
- S. U. Noble, Algorithms of Oppression: How Search Engines Reinforce Racism. NYU Press, 2018.Google ScholarCross Ref
- C. O'Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. New York: Crown, 2016.Google ScholarDigital Library
- F. Pasquale, The Black Box Society: The Secret Algorithms That Control Money and Information. Harvard University Press, 2015.Google ScholarCross Ref
- J. Buolamwini and T. Gebru, ‘Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification’, in Proceedings of the 1st Conference on Fairness, Accountability and Transparency, PMLR, Jan. 2018, pp. 77–91. Accessed: May 12, 2023. [Online]. Available: https://proceedings.mlr.press/v81/buolamwini18a.htmlGoogle Scholar
- M. K. Scheuerman, M. Pape, and A. Hanna, ‘Auto-essentialization: Gender in automated facial analysis as extended colonial project’, Big Data & Society, vol. 8, no. 2, p. 20539517211053710, Jul. 2021, doi: 10.1177/20539517211053712.Google ScholarCross Ref
- K. Crawford, The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press, 2021.Google Scholar
- B. Laufer, S. Jain, A. F. Cooper, J. Kleinberg, and H. Heidari, ‘Four Years of FAccT: A Reflexive, Mixed-Methods Analysis of Research Contributions, Shortcomings, and Future Prospects’, in 2022 ACM Conference on Fairness, Accountability, and Transparency, in FAccT ’22. New York, NY, USA: Association for Computing Machinery, Jun. 2022, pp. 401–426. doi: 10.1145/3531146.3533107.Google ScholarDigital Library
- L. Larue and T. M. Mueller, ‘La Normativité en Science Economique. Une perspective pratique, historique et philosophique’, Revue Philosophique de Louvain, vol. 116, p. 147, 2018.Google Scholar
- T. M. Porter, Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton University Press, 1996.Google ScholarCross Ref
- È. Chiapello and A. Desrosières, ‘18. La quantification de l’économie et la recherche en sciences sociales : paradoxes, contradictions et omissions. Le cas exemplaire de la positive accounting theory’, in L’économie des conventions, méthodes et résultats, in Recherches. Paris: La Découverte, 2006, pp. 297–310. doi: 10.3917/dec.eymar.2006.01.0297.Google Scholar
- A. Desrosières, Pour une sociologie historique de la quantification: L'Argument statistique I. Presses des Mines via OpenEdition, 2013.Google Scholar
- O. Martin, L'empire des chiffres: Sociologie de la quantification. Malakoff, 2020.Google Scholar
- L. Daston and P. Galison, Objectivity. Zone Books, 2010.Google Scholar
- G. Bachelard, Le Nouvel Esprit Scientifique. Presses Universitaires de France, 1934.Google Scholar
- B. Friedman and D. G. Hendry, Value Sensitive Design: Shaping Technology with Moral Imagination. MIT Press, 2019.Google ScholarCross Ref
- H. Nissenbaum, ‘How computer systems embody values’, Computer, vol. 34, no. 3, pp. 120–119, Mar. 2001, doi: 10.1109/2.910905.Google ScholarDigital Library
- M. Kearns and A. Roth, The Ethical Algorithm: The Science of Socially Aware Algorithm Design. Oxford University Press, 2019.Google Scholar
- L. Introna and H. Nissenbaum, ‘Defining the Web: the politics of search engines’, Computer, vol. 33, no. 1, pp. 54–62, Jan. 2000, doi: 10.1109/2.816269.Google ScholarDigital Library
- B. Friedman and H. Nissenbaum, ‘Bias in computer systems’, ACM Trans. Inf. Syst., vol. 14, no. 3, pp. 330–347, Jul. 1996, doi: 10.1145/230538.230561.Google ScholarDigital Library
- C. Dwork, F. McSherry, K. Nissim, and A. Smith, ‘Calibrating Noise to Sensitivity in Private Data Analysis’, in Theory of Cryptography, S. Halevi and T. Rabin, Eds., in Lecture Notes in Computer Science. Berlin, Heidelberg: Springer, 2006, pp. 265–284. doi: 10.1007/11681878_14.Google ScholarDigital Library
- C. Dwork and D. K. Mulligan, ‘It's Not Privacy, and It's Not Fair’, Stan. L. Rev. Online, vol. 66, p. 35, 2014 2013.Google Scholar
- L. Sweeney, ‘Discrimination in Online Ad Delivery’. arXiv, Jan. 28, 2013. doi: 10.48550/arXiv.1301.6822.Google Scholar
- L. Code, Epistemic Responsibility. State University of New York Press, 2020.Google Scholar
- J.-M. John-Mathews and D. Cardon, ‘The Crisis of Social Categories in the Age of AI’, Sociologica, vol. 16, no. 3, Art. no. 3, 2022, doi: 10.6092/issn.1971-8853/15931.Google Scholar
- L. Breiman, ‘Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author)’, Statist. Sci., vol. 16, no. 3, pp. 199–231, Aug. 2001, doi: 10.1214/ss/1009213726.Google ScholarCross Ref
- M. K. Scheuerman, A. Hanna, and E. Denton, ‘Do Datasets Have Politics? Disciplinary Values in Computer Vision Dataset Development’, Proc. ACM Hum.-Comput. Interact., vol. 5, no. CSCW2, p. 317:1-317:37, Oct. 2021, doi: 10.1145/3476058.Google ScholarDigital Library
- G. Bachelard, Le rationalisme appliqué. Presses Universitaires de France, 1949.Google Scholar
- A. Birhane , ‘The Forgotten Margins of AI Ethics’, in 2022 ACM Conference on Fairness, Accountability, and Transparency, in FAccT ’22. New York, NY, USA: Association for Computing Machinery, Jun. 2022, pp. 948–958. doi: 10.1145/3531146.3533157.Google ScholarDigital Library
- B. Hutchinson and M. Mitchell, ‘50 Years of Test (Un)fairness: Lessons for Machine Learning’, in Proceedings of the Conference on Fairness, Accountability, and Transparency, in FAT* ’19. New York, NY, USA: Association for Computing Machinery, Jan. 2019, pp. 49–58. doi: 10.1145/3287560.3287600.Google ScholarDigital Library
- J. Goldenfein, ‘The Profiling Potential of Computer Vision and the Challenge of Computational Empiricism’, in Proceedings of the Conference on Fairness, Accountability, and Transparency, in FAT* ’19. New York, NY, USA: Association for Computing Machinery, Jan. 2019, pp. 110–119. doi: 10.1145/3287560.3287568.Google ScholarDigital Library
- D. Haraway, ‘Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective’, Feminist Studies, vol. 14, no. 3, pp. 575–599, 1988, doi: 10.2307/3178066.Google ScholarCross Ref
- F. Jaton, ‘Assessing biases, relaxing moralism: On ground-truthing practices in machine learning design and application’, Big Data & Society, vol. 8, no. 1, p. 20539517211013570, Jan. 2021, doi: 10.1177/20539517211013569.Google ScholarCross Ref
- A. Desrosières, The Politics of Large Numbers: A History of Statistical Reasoning. Harvard University Press, 2002.Google Scholar
- P. J. Brantingham, ‘The Logic of Data Bias and its Impact on Place-Based Predictive Policing’, Ohio St. J. Crim. L., vol. 15, p. 473, 2018 2017.Google Scholar
- N. Mehrabi, F. Morstatter, N. Saxena, K. Lerman, and A. Galstyan, ‘A Survey on Bias and Fairness in Machine Learning’, ACM Comput. Surv., vol. 54, no. 6, p. 115:1-115:35, Jul. 2021, doi: 10.1145/3457607.Google ScholarDigital Library
- J. Kleinberg, J. Ludwig, and S. Mullainathan, ‘A Guide to Solving Social Problems with Machine Learning’, Harvard Business Review, Dec. 08, 2016. https://hbr.org/2016/12/a-guide-to-solving-social-problems-with-machine-learning (accessed Sep. 09, 2018).Google Scholar
- R. Abebe, S. Barocas, J. Kleinberg, K. Levy, M. Raghavan, and D. G. Robinson, ‘Roles for computing in social change’, in Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, in FAT* ’20. New York, NY, USA: Association for Computing Machinery, Jan. 2020, pp. 252–260. doi: 10.1145/3351095.3372871.Google ScholarDigital Library
- A. Desrosières, Prouver et gouverner: Une analyse politique des statistiques publiques. La Découverte, 2014.Google ScholarCross Ref
- A. Castelnovo, R. Crupi, G. Greco, D. Regoli, I. G. Penco, and A. C. Cosentini, ‘A clarification of the nuances in the fairness metrics landscape’, Sci Rep, vol. 12, no. 1, Art. no. 1, Mar. 2022, doi: 10.1038/s41598-022-07939-1.Google Scholar
- B. Latour, Cogitamus: six lettres sur les humanités scientifiques. La Découverte, 2014.Google Scholar
- J. Kleinberg, S. Mullainathan, and M. Raghavan, ‘Inherent Trade-Offs in the Fair Determination of Risk Scores’, in 8th Innovations in Theoretical Computer Science Conference (ITCS 2017), C. H. Papadimitriou, Ed., in Leibniz International Proceedings in Informatics (LIPIcs), vol. 67. Dagstuhl, Germany: Schloss Dagstuhl–Leibniz-Zentrum fuer Informatik, 2017, p. 43:1-43:23. doi: 10.4230/LIPIcs.ITCS.2017.43.Google Scholar
- M. Hardt, E. Price, E. Price, and N. Srebro, ‘Equality of Opportunity in Supervised Learning’, in Advances in Neural Information Processing Systems, Curran Associates, Inc., 2016. Accessed: Dec. 14, 2022. [Online]. Available: https://proceedings.neurips.cc/paper/2016/hash/9d2682367c3935defcb1f9e247a97c0d-Abstract.htmlGoogle Scholar
- L. Daston, ‘The Moral Economy of Science’, Osiris, vol. 10, pp. 2–24, Jan. 1995.Google ScholarCross Ref
- T. Kirat, O. Tambou, V. Do, and A. Tsoukiàs, ‘Fairness and Explainability in Automatic Decision-Making Systems. A challenge for computer science and law’. arXiv, May 13, 2022. doi: 10.48550/arXiv.2206.03226.Google Scholar
- S. Wachter, B. Mittelstadt, and C. Russell, ‘Why fairness cannot be automated: Bridging the gap between EU non-discrimination law and AI’, Computer Law & Security Review, vol. 41, p. 105567, Jul. 2021, doi: 10.1016/j.clsr.2021.105567.Google ScholarCross Ref
- M. Kearns, S. Neel, A. Roth, and Z. S. Wu, ‘Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness’. arXiv, Dec. 03, 2018. doi: 10.48550/arXiv.1711.05144.Google Scholar
- C. Dwork, M. Hardt, T. Pitassi, O. Reingold, and R. Zemel, ‘Fairness through awareness’, in Proceedings of the 3rd Innovations in Theoretical Computer Science Conference, in ITCS ’12. New York, NY, USA: Association for Computing Machinery, Jan. 2012, pp. 214–226. doi: 10.1145/2090236.2090255.Google ScholarDigital Library
- M. Fourcade, ‘Ordinalization: Lewis A. Coser memorial award for theoretical agenda setting 2014’, Sociological Theory, vol. 34, no. 3, pp. 175–195, 2016.Google ScholarCross Ref
- J.-M. John-Mathews, ‘L’Éthique de l'Intelligence Artificielle en Pratique. Enjeux et Limites’, These de doctorat, université Paris-Saclay, 2021. Accessed: May 11, 2023. [Online]. Available: https://www.theses.fr/2021UPASI015Google Scholar
- M. J. Kusner, C. Russell, J. R. Loftus, and R. Silva, ‘Causal Interventions for Fairness’, arXiv:1806.02380 [cs, stat], Jun. 2018, Accessed: Jan. 13, 2021. [Online]. Available: http://arxiv.org/abs/1806.02380Google Scholar
- V. Tremblay, ‘Algorithmic fairness: interdisciplinary perspective and recommendations for statisticians and other data scientists’. May 10, 2022. Accessed: Jan. 11, 2023. [Online]. Available: https://hal.science/hal-03663226Google Scholar
- K. Makhlouf, S. Zhioua, and C. Palamidessi, ‘Survey on Causal-based Machine Learning Fairness Notions’, arXiv:2010.09553 [cs], Jan. 2021, Accessed: Feb. 07, 2021. [Online]. Available: http://arxiv.org/abs/2010.09553Google Scholar
- C. Tiercelin, Le ciment des choses, 1er édition. Paris: Editions Ithaque, 2011.Google Scholar
- L. Hu and I. Kohler-Hausmann, ‘What's Sex Got To Do With Fair Machine Learning?’, in Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, Jan. 2020, pp. 513–513. doi: 10.1145/3351095.3375674.Google ScholarDigital Library
- L. Boltanski and L. Thévenot, De la justification: les économies de la grandeur. Gallimard, 1991.Google Scholar
- F. Nef and S. Berlioz, La nature du social: de quoi le social est-il fait? Le Bord de l'eau, 2021.Google Scholar
- J.-M. John-Mathews, D. Cardon, and C. Balagué, ‘From Reality to World. A Critical Perspective on AI Fairness’, J Bus Ethics, vol. 178, no. 4, pp. 945–959, Jul. 2022, doi: 10.1007/s10551-022-05055-8.Google ScholarCross Ref
Index Terms
- Fairness in machine learning from the perspective of sociology of statistics: How machine learning is becoming scientific by turning its back on metrological realism
Recommendations
Algorithmic Fairness from a Non-ideal Perspective
AIES '20: Proceedings of the AAAI/ACM Conference on AI, Ethics, and SocietyInspired by recent breakthroughs in predictive modeling, practitioners in both industry and government have turned to machine learning with hopes of operationalizing predictions to drive automated decisions. Unfortunately, many social desiderata ...
From daguerreotypes to algorithms: machines, expertise, and three forms of objectivity
What claims are made about the objectivity of machines versus that of human experts? Whereas most current debates focus on the growing impact of algorithms in the age of Big Data, I argue here in favor of taking a longer historical perspective on these ...
Information, analysis, and ideology: A case study of science and the public interest
The politicization of science is not a new phenomenon, but the disputes surrounding global climate change have been particularly subject to ideological positioning. The work conducted by researchers on the description of, and possible causes for, ...
Comments