Measuring Fairness in Ranked Outputs

By Ke Yang, Julia Stoyanovich. 2016

378 citations recorded by the source corpus.

Ranking and scoring are ubiquitous. We consider the setting in which an institution, called a ranker, evaluates a set of individuals based on demographic, behavioral or other characteristics. The final output is a ranking that represents the relative quality of the individuals. While automatic and therefore seemingly objective, rankers can, and often do, discriminate against individuals and systematically disadvantage members of protected groups. This warrants a careful study of the fairness of a ranking scheme, to enable data science for social good applications, among others. In this paper we propose fairness measures for ranked outputs. We develop a data generation procedure that allows us to systematically control the degree of unfairness in the output, and study the behavior of our measures on these datasets. We then apply our proposed measures to several real datasets, and detect cases of bias. Finally, we show preliminary results of incorporating our ranked fairness measures into an optimization framework, and show potential for improving fairness of ranked outputs while maintaining accuracy. The code implementing all parts of this work is publicly available at https://github.com/DataResponsibly/FairRank.

Subfields: Artificial Intelligence, Safety Research

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

In Safety Research

Citations in the corpus

FairSight: Visual Analytics for Fairness in Decision Making (2019, background): “The fairness problem here is to pursue the degree of exposure and attention fairly for groups [37, 39] or individuals [4].”
Stable and Fair Classification (2019, background): “, ranking [18, 76], summarization [17], personalization [19, 16], multiwinner voting [15], and online advertising [13].”
A Nutritional Label for Rankings (2018, methodology): “The Fairness widget quanti es whether the ranked output exhibits statistical parity (one interpretation of fairness) with respect to one or more sensitive attributes, such as gender or race of individuals [13].”
Fairness in Machine Learning (2020, methodology): “…21, 22, 23, 25, 28, 29, 31, 34, 44, 45, 46, 52, 53, 54, 55, 56, 59, 66, 69, 71, 72, 73, 74, 75, 82, 84, 87, 89, 90, 94, 95, 97, 98, 99, 103, 104, 106, 114, 115, 116, 121, 123, 126, 138, 139, 141, 148, 149, 153, 156, 159, 165, 174, 176, 177, 178, 179, 180, 182, 183, 184, 186, 187, 191, 192, 193].”
Anti-discrimination learning: a causal modeling-based framework (2017, background): “Some preliminary works (e.g., [36]) have proposed several association-based methods.”
Fair Algorithms for Clustering (2019, background): “A considerable amount of work [14,18,30,31,42–44] addressing this issue has emerged in the recent years.”
Interpretability of deep learning models: A survey of results (2017, background): “[37] consider the problem of ranking a set of individuals bvased on demographic, behavioral or other characteristics wherein rankers can, and often do, discriminate against individuals and disadvantaged members of protected groups despite seemingly automatic and objective metrics [38].”
Diversity and Inclusion Metrics in Subset Selection (2020, background): “Some notions of fairness proposed in the ranking and subset selection literature include considerations that are closely related to the idea of diversity discussed here [2, 8, 13, 32, 36].”
Fair and Diverse DPP-based Data Summarization (2018, methodology): “th 14 features such as age, race, education and a binary label indicating whether a subject’s incomes is above or below 50K USD.3 This dataset has been widely studied in the context of fairness (see, [27, 29, 30, 28]). In preprocessing the data we lter out incomplete entries, and from the remaining ”
Using Image Fairness Representations in Diversity-Based Re-ranking for Recommendations (2018, background): “Lastly, a few fairness measures for ranked outputs have been proposed in Yang and Stoyanovich [19] and we intend to explore the performance of FMMR further using these measures.”
Designing Fair Ranking Schemes (2017, methodology): “Several recent papers focus on measuring fairness in ranked lists [31,32], on constructing ranked lists that meet fairness criteria [9], and on fair and diverse set selection [27].”
A Survey on the Fairness of Recommender Systems (2022, methodology): “For items, some researchers [31, 102] consider unfairness across different positions and then aggregate them.”
Fairness of Exposure in Rankings (2018, background): “Yang and Stoyanovich [30] propose statistical parity based measures that compute the difference in the distribution of different groups for different prefixes of the ranking (top-10, top20 and so on).”
Classification with Fairness Constraints: A Meta-Algorithm with Provable Guarantees (2018, background): “We also note that our linear fairness constraints are inspired by works on other fundamental algorithmic problems such as data summarization [11], ranking [12, 58], elections [10], and personalization [13].”