Inherent Trade-Offs in the Fair Determination of Risk Scores

@article{Kleinberg2017InherentTI,
  title={Inherent Trade-Offs in the Fair Determination of Risk Scores},
  author={Jon M. Kleinberg and Sendhil Mullainathan and M. Raghavan},
  journal={ArXiv},
  year={2017},
  volume={abs/1609.05807}
}
Recent discussion in the public sphere about algorithmic classification has involved tension between competing notions of what it means for a probabilistic classification to be fair to different groups. [] Key Result These results suggest some of the ways in which key notions of fairness are incompatible with each other, and hence provide a framework for thinking about the trade-offs between them.
Inherent Trade-Offs in Algorithmic Fairness
TLDR
Several of the key fairness conditions that lie at the heart of debates about classification by algorithms are considered, and recent research establishes inherent trade-offs between these conditions.
Inherent Trade-Offs in Algorithmic Fairness
TLDR
Several of the key fairness conditions that lie at the heart of debates, and recent research establishing inherent trade-offs between these conditions are considered, are considered.
FACT: A Diagnostic for Group Fairness Trade-offs
TLDR
A general diagnostic is proposed that enables systematic characterization of multiple trade-offs including group fairness incompatibilities and suggests an alternate post-processing method for designing fair classifiers.
A Possibility in Algorithmic Fairness: Calibrated Scores for Fair Classifications
TLDR
This paper derives necessary and sufficient conditions for the existence of calibrated scores that yield classifications achieving equal error rates, and presents an algorithm that searches for the most informative score subject to both calibration and minimal error rate disparity.
On statistical criteria of algorithmic fairness
TLDR
It is argued that none of the purported criteria, except for a calibration criterion, are necessary conditions for fairness, on the grounds that they can all be simultaneously violated by a manifestly fair and uniquely optimal predictive algorithm, even when base rates are equal.
On the Choice of Fairness: Finding Representative Fairness Metrics for a Given Context
TLDR
A framework that automatically discovers the correlations and trade-offs between different pairs of measures for a given context and allows users to view unfairness from various perspectives that might otherwise be ignored due to the sheer size of the exploration space is proposed.
How Do the Score Distributions of Subpopulations Influence Fairness Notions?
TLDR
It is shown in this study that some profit-fairness trade-offs and fairness-fairnesses depend substantially on the underlying score distributions given to subpopulations and it is shown that higher symmetry in scores of sub Populations can significantly reduce the trade-off between fairness notions within a given acceptable strictness, even when sacrificing expressiveness.
From Parity to Preference-based Notions of Fairness in Classification
TLDR
This paper draws inspiration from the fair-division and envy-freeness literature in economics and game theory and proposes preference-based notions of fairness -- any group of users would collectively prefer its treatment or outcomes, regardless of the (dis)parity as compared to the other groups.
A clarification of the nuances in the fairness metrics landscape
TLDR
This work tries to make some order out of this zoo of definitions of fairness in ML, that consider different notions of what is a “fair decision” in situations impacting individuals in the population.
A Resolution in Algorithmic Fairness: Calibrated Scores for Fair Classifications
TLDR
This paper derives necessary and sufficient conditions for the existence of calibrated scores that yield classifications achieving equal error rates and presents an algorithm that searches for the most informative score subject to both calibration and minimal error rate disparity.
...
...

References

SHOWING 1-10 OF 50 REFERENCES
Fairness in Learning: Classic and Contextual Bandits
TLDR
A tight connection between fairness and the KWIK (Knows What It Knows) learning model is proved: a provably fair algorithm for the linear contextual bandit problem with a polynomial dependence on the dimension, and a worst-case exponential gap in regret between fair and non-fair learning algorithms.
Algorithmic Decision Making and the Cost of Fairness
TLDR
This work reformulate algorithmic fairness as constrained optimization: the objective is to maximize public safety while satisfying formal fairness constraints designed to reduce racial disparities, and also to human decision makers carrying out structured decision rules.
Fairness Constraints: Mechanisms for Fair Classification
TLDR
This paper introduces a flexible mechanism to design fair classifiers by leveraging a novel intuitive measure of decision boundary (un)fairness, and shows on real-world data that this mechanism allows for a fine-grained control on the degree of fairness, often at a small cost in terms of accuracy.
Fairness through awareness
TLDR
A framework for fair classification comprising a (hypothetical) task-specific metric for determining the degree to which individuals are similar with respect to the classification task at hand and an algorithm for maximizing utility subject to the fairness constraint, that similar individuals are treated similarly is presented.
Fairness-aware Learning through Regularization Approach
TLDR
This paper discusses three causes of unfairness in machine learning and proposes a regularization approach that is applicable to any prediction algorithm with probabilistic discriminative models and applies it to logistic regression to empirically show its effectiveness and efficiency.
Learning Fair Representations
We propose a learning algorithm for fair classification that achieves both group fairness (the proportion of members in a protected group receiving positive classification is identical to the
Fair prediction with disparate impact: A study of bias in recidivism prediction instruments
TLDR
It is demonstrated that the criteria cannot all be simultaneously satisfied when recidivism prevalence differs across groups, and how disparate impact can arise when an RPI fails to satisfy the criterion of error rate balance.
Certifying and Removing Disparate Impact
TLDR
This work links disparate impact to a measure of classification accuracy that while known, has received relatively little attention and proposes a test for disparate impact based on how well the protected class can be predicted from the other attributes.
Equality of Opportunity in Supervised Learning
TLDR
This work proposes a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features and shows how to optimally adjust any learned predictor so as to remove discrimination according to this definition.
Learning Fair Classifiers
TLDR
This paper introduces a flexible mechanism to design fair classifiers in a principled manner and instantiates this mechanism on three well-known classifiers -- logistic regression, hinge loss and linear and nonlinear support vector machines.
...
...