A Reductions Approach to Fair Classification
By Alekh Agarwal, Alina Beygelzimer, Miroslav Dud'ik, J. Langford, Hanna M. Wallach. 2018
1,292 citations recorded by the source corpus.
We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sensitive classification problems, whose solutions yield a randomized classifier with the lowest (empirical) error subject to the desired constraints. We introduce two reductions that work for any representation of the cost-sensitive classifier and compare favorably to prior baselines on a variety of data sets, while overcoming several of their disadvantages.
Subfields: Artificial Intelligence, Safety Research, AI Ethics, Fairness & Society
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Percy Liang
- Sergey Levine
- Yoshua Bengio
- Mohit Bansal
- Ruslan Salakhutdinov
- Quoc V. Le
- Jianfeng Gao
- Luke S. Zettlemoyer
In Safety Research
- Silvia Chiappa
- Dana Pessach
- Erez Shmueli
- Luca Oneto
- Aaron Roth
- Alexandra Chouldechova
- Adrian Weller
- Heinrich Jiang
In AI Ethics, Fairness & Society
- Dana Pessach
- Erez Shmueli
- Luca Oneto
- Silvia Chiappa
- Christian Haas
- Simon Caton
- Aaron Roth
- Berk Ustun
Citations in the corpus
- Beyond Individual and Group Fairness (2020, background): “…of recent research work on the topic of fairness and algorithmic bias in machine learning [Dwork et al., 2012, Zemel et al., 2013, Hardt et al., 2016, Kleinberg et al., 2017, Pleiss et al., 2017, Agarwal et al., 2018, Kearns et al., 2018, Gillen et al., 2018, Sharifi-Malvajerdi et al., 2019].”
- Differential Privacy Has Disparate Impact on Model Accuracy (2019, background): “Recent research [10, 20] aims to add fairness and DP to post-processing [17] and in-processing [2] algorithms.”
- Fair Resource Allocation in Federated Learning (2019, background): “Another set of works optimize an objective subject to some fairness constraints during training time [3, 6, 18, 43, 46, 47, 9].”
- Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness (2017, methodology): “Technically, the most closely related piece of work (and from which we take inspiration for our algorithm in Section 4) is (Agarwal et al., 2017), who show that given access to an agnostic learning oracle for a class H, there is an efficient algorithm to find the lowest-error distribution over class”
- Distill-and-Compare: Auditing Black-Box Models Using Transparent Model Distillation (2017, background): “Other approaches also train their own outcome models, then uncover feature regions where the model is not accurate [3, 23, 24, 38].”
- Statistical Equity: A Fairness Classification Objective (2020, background): “…also targets a wide variety of fairness definitions, such as equality of opportunity and equalized odds [Hardt et al., 2016, Woodworth et al., 2017], statistical parity [Agarwal et al., 2018], and subgroup fairness [Ustun, Liu, and Parkes, 2019], and their incorporation in classification task.”
- Wasserstein Fair Classification (2019, methodology): “Some methods transform the constrained optimization problem via the method of Lagrange multipliers (Agarwal et al., 2018; Corbett-Davies et al., 2017; Cotter et al., 2018; Goh et al., 2016; Narasimhan, 2018; Wu et al., 2018; Zafar et al., 2017).”
- Fair Regression: Quantitative Definitions and Reduction-based Algorithms (2019, methodology): “Similarly to our analysis of Algorithm 1 in Theorem 4, we can appeal to Theorem 1, and Lemmas 2 and 3 of Agarwal et al. (2018).”
- Stable and Fair Classification (2019, methodology): “Given a classifier f , we consider a fairness metric for statistical rate, which has been applied in [55, 4].”
- Fairness and Abstraction in Sociotechnical Systems (2019, background): “portable design: the entire paper describes a process for building a "fair wrapper" around any classifier to make the resulting outputs fair [4].”
- Taking Advantage of Multitask Learning for Fair Classification (2018, background): “In many recent works [2–4 8 9 16 17 17 25 26 28 30 35–38] it has been shown how to enforce these constraints during the learning phase of a classifier.”
- An Empirical Study of Rich Subgroup Fairness for Machine Learning (2018, background): “Finally, perhaps enforcing combinatorially many constraints is not necessary — perhaps on real data, it is enough to call upon the algorithm of [1] for enforcing statistical fairness constraints on the small number of groups defined by the marginal protected attributes, and rich subgroup fairness wi”
- Characterizing Fairness Over the Set of Good Models Under Selective Labels (2021, background): “Fairlearn searches for the lossminimizing model subject to a fairness parity constraint (Agarwal et al., 2018).”
- Improving Fairness in Machine Learning Systems: What Do Industry Practitioners Need? (2018, background): “Significant effort in the fair ML community has focused on the development of statistical definitions of fairness [14, 25, 32, 83] and algorithmic methods to assess and mitigate biases in relation to these definitions [3, 17, 49, 68].”