Fairness through awareness

  title={Fairness through awareness},
  author={Cynthia Dwork and Moritz Hardt and Toniann Pitassi and Omer Reingold and Richard S. Zemel},
We study fairness in classification, where individuals are classified, e.g., admitted to a university, and the goal is to prevent discrimination against individuals based on their membership in some group, while maintaining utility for the classifier (the university). [] Key Method The main conceptual contribution of this paper is a framework for fair classification comprising (1) a (hypothetical) task-specific metric for determining the degree to which individuals are similar with respect to the…

Figures from this paper

Temporal Aspects of Individual Fairness

Two definitions of fairness-across-time can have drastically different implications in the setting where the principal needs to learn the utility model: one can achieve a vanishing asymptotic loss in long-run average utility relative to the full-information optimum under the fairness-in-hindsight constraint, whereas this asymPTotic loss can be bounded away from zero under the fair-ACross- time constraint.

Metric Learning for Individual Fairness

This work proposes a solution to the problem of approximating a metric for Individual Fairness based on human judgments by assuming that the arbiter can answer a limited set of queries concerning similarity of individuals for a particular task, is free of explicit biases and possesses sufficient domain knowledge to evaluate similarity.

On the Compatibility of Privacy and Fairness

This work investigates whether privacy and fairness can be simultaneously achieved by a single classifier in several different models and gives an efficient algorithm for classification that maintains utility and satisfies both privacy and approximate fairness with high probability.

Eliciting and Enforcing Subjective Individual Fairness

A framework for fairness elicitation is considered, in which fairness is indirectly specified only via a sample of pairs of individuals who should be treated (approximately) equally on the task, and a provably convergent oracle-efficient algorithm is provided for minimizing error subject to the fairness constraints.

What's Fair about Individual Fairness?

It is suggested that individual fairness cannot be a definition of fairness, and instead should be seen as one tool among several for ameliorating algorithmic bias.

Fair for All: Best-effort Fairness Guarantees for Classification

A fairness notion whose guarantee, on each group g in a class G, is relative to the performance of the best classifier on g is proposed, which applies to broad classes of groups, in particular, where G consists of all possible groups (subsets) in the data, and G is more streamlined.

On the Applicability of ML Fairness Notions

This paper is a survey of fairness notions that addresses the question of "which notion of fairness is most suited to a given real-world scenario and why?".

Distributional Individual Fairness in Clustering

This paper adopts the individual fairness notion, which mandates that similar individuals should be treated similarly for clustering problems, and introduces a framework for assigning individuals, embedded in a metric space, to probability distributions over a bounded number of cluster centers.

Individual Fairness in Hindsight

Fairness-across-time and fairness-in-hindsight are introduced and a new algorithm is designed, Cautious Fair Exploration (CAFE), which satisfies FH and achieves sub-linear regret guarantees for a broad range of settings.

The cost of fairness in binary classification

This work relates two existing fairness measures to cost-sensitive risks, and shows that for such costsensitive fairness measures, the optimal classifier is an instance-dependent thresholding of the class-probability function.



Mechanism Design via Differential Privacy

It is shown that the recent notion of differential privacv, in addition to its own intrinsic virtue, can ensure that participants have limited effect on the outcome of the mechanism, and as a consequence have limited incentive to lie.

On allocations that maximize fairness

This work considers a problem known as the restricted assignment version of the max-min allocation problem with indivisible goods, and presents a rounding technique that recovers an allocation of value at least Ω(log log log m/log log m) of the optimum.

Incorporating Fairness into Game Theory and Economics

People like to help those who are helping them and to hurt those who are hurting them. Outcomes rejecting such motivations are called fairness equilibria. Outcomes are mutual-max when each person

Mechanism design with uncertain inputs: (to err is human, to forgive divine)

A task of scheduling with a common deadline on a single machine is considered, where players are uncertain about their own job lengths, and hence are incapable of providing truthful reports, and a probabilistic model for uncertainty is shown that even with relatively little uncertainty, no mechanism can guarantee a constant fraction of the maximum welfare.

Differential Privacy

  • C. Dwork
  • Computer Science
    Encyclopedia of Cryptography and Security
  • 2006
A general impossibility result is given showing that a formalization of Dalenius' goal along the lines of semantic security cannot be achieved, which suggests a new measure, differential privacy, which, intuitively, captures the increased risk to one's privacy incurred by participating in a database.

Calibrating Noise to Sensitivity in Private Data Analysis

The study is extended to general functions f, proving that privacy can be preserved by calibrating the standard deviation of the noise according to the sensitivity of the function f, which is the amount that any single argument to f can change its output.

Fairness in scheduling

This paper explores the issue of fairness in scheduling in settings where there are long-lived processes which should be repeatedly scheduled for various tasks throughout the lifetime of a system, and develops a notion of desired load of a process, which is a function of the tasks it participates in.

Gaydar: Facebook Friendships Expose Sexual Orientation

It is determined that the percentage of a given user's friends who self-identify as gay male is strongly correlated with the sexual orientation of that user, and a logistic regression classifier with strong predictive power is developed.

On the geometry of differential privacy

The lower bound is strong enough to separate the concept of differential privacy from the notion of approximate differential privacy where an upper bound of O(√{d}/ε) can be achieved.

Decentralizing equality of opportunity and issues concerning the equality of educational opportunity

Decentralizing Equality of Opportunity and Issues Concerning the Equality of Educational Opportunity Caterina Calsamiglia 2005 This dissertation explores issues concerning local versus global