计算机科学 ›› 2024, Vol. 51 ›› Issue (1): 266-272.doi: 10.11896/jsjkx.230500224
张文琼, 李云
ZHANG Wenqiong, LI Yun
摘要: 随着机器学习应用的日益普及,机器学习公平性问题引起了学术界和工业界的广泛关注,成为了可信人工智能的重要组成部分。为了评估和改善机器学习应用的公平性,研究人员提出了一系列公平性指标,这些指标有助于保障机器学习模型在不同个体、群体间的公平决策,并为改善和优化模型提供指导。但各界对于指标之间的区别与联系仍没有形成共识,对不同场景、不同任务的公平性定义没有明确的划分,公平性指标缺乏完善的分类体系。文中对公平性指标进行了全面的整理和归类,从指标的数学定义出发,根据是否基于概率统计将公平性指标分为两类,然后分别对这两类指标进行进一步的细粒度划分和阐述。为了便于读者理解和运用,结合一个实际案例,从适用场景和实现条件等方面指出各类指标的优势和面临的挑战,还结合数学定义讨论了指标之间的关系,并对未来趋势进行了展望。
中图分类号:
[1]MAHAJAN S,CARABALLO C,LU Y,et al.Trends in diffe-rences in health status and health care access and affordability by race and ethnicity in the United States,1999-2018[J].Jama,2021,326(7):637-648. [2]CAUFFMAN C.Discrimination in online advertising[J].Maastricht Journal of European and Comparative Law,2021,28(3):283-286. [3]LEE M S,FLORIDI L.Algorithmic fairness in mortgage len-ding:from absolute conditions to relational trade-offs[J].Minds and Machines,2021,31(1):165-191. [4]XU G N,CHEN Y P,CHEN Y M,et al.Data Debiasing Method Based on Constrained Optimized Generative Adversarial Networks[J].Computer Science,2022,49(6A):184-190. [5]PEDRESHI D,RUGGIERI S,TURINI F.Discrimination-aware data mining[C]//International Conference on Knowledge Discovery and Data Mining.ACM,2008:560-568. [6]VERMA S,RUBIN J.Fairness definitions explained[C]//International Workshop on Software Fairness(Fairware).ACM,2018:1-7. [7]MEHRABI N,MORSTATTER F,SAXENA N,et al.A survey on bias and fairness in machine learning[J].ACM Computing Surveys,2021,54(6):1-35. [8]BERK R,HEIDARI H,JABBARI S,et al.Fairness in criminal justice risk assessments:The state of the art[J].Sociological Methods Research,2021,50(1):3-44. [9]ZAFAR M B,VALERA I,GOMEZ R M,et al.Fairness beyond disparate treatment disparate impact:learning classification without disparate mistreatment[C]//International Conference on World Wide Web.ACM,2017:1171-1180. [10]HARDT M,PRICE E,SREBRO N.Equality of opportunity in supervised learning[J].Advances in Neural Information Processing Systems,2016,29(1):3323-3331. [11]CALDERS T,VERWER S.Three naive Bayes approaches for discrimination-free classification[J].Data Mining and Know-ledge Discovery,2010,21(2):277-292. [12]JIANG Z,HAN X,FAN C,et al.Generalized demographic parity for group fairness[C]//International Conference on Learning Representations.OpenReview.net,2022. [13]KAMIRAN F,LIOBAITÈ I,CALDERS T.Quantifying ex-plainable discrimination and removing illegal discrimination in automated decision making[J].Knowledge and Information Systems,2013,35(3):613-644. [14]XU R,CUI P,KUANG K,et al.Algorithmic decision making with conditional fairness[C]//International Conference on Knowledge Discovery and Data Mining.ACM,2020:2125-2135. [15]FELDMAN M,FRIEDLER S A,MOELLER J,et al.Certifying and removing disparate impact[C]//International Conference on Knowledge Discovery and Data Mining.ACM,2015:259-268. [16]STEWART R T.Identity and the limits of fair assessment[J].Journal of Theoretical Politics,2022,34(3):415-442. [17]HEDDEN B.On statistical criteria of algorithmic fairness[J].Philosophy and Public Affairs,2021,49(2):209-231. [18]CORBETT D S,PIERSON E,FELLER A,et al.Algorithmic decision making and the cost of fairness[C]//International Conference on Knowledge Discovery and Data Mining.ACM,2017:797-806. [19]ZHAO H,GORDON G.Inherent tradeoffs in learning fair representations[J].Machine Learning Research,2022,23(57):1-26. [20]DEHO O B,ZHAN C,LI J,et al.How do the existing fairness metrics and unfairness mitigation algorithms contribute to ethical learning analytics?[J].British Journal of Educational Technology,2022,53(4):822-843. [21]BAROCAS S,SELBST A D.Big data’s disparate impact[J].Calif.L.Rev.,2016,104:671-732. [22]CASTELNOVO A,CRUPI R,GRECO G,et al.A clarification of the nuances in the fairness metrics landscape[J].Scientific Reports,2022,12(1):4209. [23]DWORK C,HARDT M,PITASSI T,et al.Fairness throughawareness[C]//International Conference on Innovations in Theoretical Computer Science.ACM,2012:214-226. [24]LI A,PEARL J.Unit selection with causal diagram[C]//International Conference on Artificial Intelligence.AAAI,2022,36(5):5765-5772. [25]GEBHART V,SMERZI A.Extending the fair sampling as-sumption using causal diagrams[J].Quantum,2023,7:897-906. [26]KILBERTUS N,ROJAS C M,PARASCANDOLO G,et al.Avoiding discrimination through causal reasoning[J].Advances in Neural Information Processing Systems,2017,30(1):656-666. [27]KIM H,SHIN S,JANG J H,et al.Counterfactual fairness with disentangled causal effect variational autoencoder[C]//International Conference on Artificial Intelligence.AAAI,2021,35(9):8128-8136. [28]KUSNER M J,LOFTUS J,RUSSELL C,et al.Counterfactual fairness[J].Advances in Neural Information Processing Systems,2017,30(1):4066-4076. [29]PUJOL D,MCKENNA R,KUPPAM S,et al.Fair decision ma-king using privacy-protected data[C]//International Conference on Fairness,Accountability,and Transparency.ACM,2020:189-199. [30]CZARNOWSKA P,VYAS Y,SHAH K.Quantifying social biases in nlp:a generalization and empirical comparison of extrinsic fairness metrics[J].Transactions of the Association for Computational Linguistics,2021,9(1):1249-1267. [31]GAO R,GE Y,SHAH C.FAIR:Fairness-aware information retrieval evaluation[J].Association for Information Science and Technology,2022,73(10):1461-1473. [32]FRIEDLER S A,SCHEIDEGGER C,VENKATASUBRAMANIAN S,et al.A comparative study of fairness-enhancing interventions in machine learning[C]//International Conference on Fairness,Accountability,and Transparency.ACM,2019:329-338. [33]KLEINBERG J,MULLAINATHAN S,RAGHAVAN M.In-herent trade-offs in the fair determination of risk scores[C]//International Conference on Innovations in Theoretical Compu-ter Science.Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik,2017:43:1-43:23. [34]LIU L T,DEAN S,ROLF E,et al.Delayed impact of fair machine learning[C]//International Conference on Machine Lear-ning.PMLR,2018:3150-3158. [35]BOUVATIER V,EL OUARDI S.Credit gaps as banking crisis predictors:a different tune for middle-and low-income countries[J].Emerging Markets Review,2023,54(C):101001. [36]FARNADI G,BABAKI B,GETOOR L.Fairness in relational domains[C]//International Conference on AI,Ethics,and Society.ACM,2018:108-114. [37]CHAO L M,YIN X L.AI Governance and System:Current Si-tuation and Trend[J].Computer Science,2021,48(9):1-8. [38]HUGGINS-MANLEY A C,BOOTH B M,D’ELLO S K.To-ward Argument-Based Fairness with an Application to AI-Enhanced Educational Assessments[J].Journal of Educational Measurement,2022,59(3):362-388. [39]GICHOYA J W,MCCOY L G,CELI L A,et al.Equity in essence:a call for operationalising fairness in machine learning for healthcare[J].BMJ Health & Care Informatics,2021,28(1):e100289. [40]KALLUS N,ZHOU A.Fairness,welfare,and equity in perso-nalized pricing[C]//Proceedings of the 2021 ACM Conference on Fairness,Accountability,and Transparency.2021:296-314. [41]BAKER R S,HAWN A.Algorithmic bias in education[J].International Journal of Artificial Intelligence in Education,2022,32(1):1052-1092. [42]REDDY S,ALLAN S,COGHLAN S,et al.A governance model for the application of AI in health care[J].Journal of the American Medical Informatics Association,2020,27(3):491-497. [43]DELOBELLE P,TOKPO E K,CALDERS T,et al.Measuring fairness with biased rulers:A comparative study on bias metrics for pre-trained language models[C]//The 2022 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies(NAACL 2022).2022:1693-1706. |
|