Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Equalizing Recourse across Groups

Vivek Gupta, Pegah Nokhiz, Chitradeep Dutta Roy, Suresh Venkatasubramanian

TL;DR

The paper tackles equalizing recourse across demographic groups by defining recourse as the distance to a classifier boundary and introducing a regularized objective to minimize disparities in this distance while maintaining accuracy. It develops an explicit, iterative framework for linear and kernelized SVMs, and a model-agnostic, LIME-based approach to approximate recourse for black-box classifiers, including a reweighting scheme to steer decision boundaries. Empirical results on multiple datasets show meaningful reductions in cross-group recourse disparity with minimal loss to predictive performance, across both SVM and model-agnostic settings. This work enables post-decision agency improvements by ensuring more uniform opportunities for individuals to change unfavorable outcomes, applicable to a range of decision-making domains.

Abstract

The rise in machine learning-assisted decision-making has led to concerns about the fairness of the decisions and techniques to mitigate problems of discrimination. If a negative decision is made about an individual (denying a loan, rejecting an application for housing, and so on) justice dictates that we be able to ask how we might change circumstances to get a favorable decision the next time. Moreover, the ability to change circumstances (a better education, improved credentials) should not be limited to only those with access to expensive resources. In other words, \emph{recourse} for negative decisions should be considered a desirable value that can be equalized across (demographically defined) groups. This paper describes how to build models that make accurate predictions while still ensuring that the penalties for a negative outcome do not disadvantage different groups disproportionately. We measure recourse as the distance of an individual from the decision boundary of a classifier. We then introduce a regularized objective to minimize the difference in recourse across groups. We explore linear settings and further extend recourse to non-linear settings as well as model-agnostic settings where the exact distance from boundary cannot be calculated. Our results show that we can successfully decrease the unfairness in recourse while maintaining classifier performance.

Equalizing Recourse across Groups

TL;DR

The paper tackles equalizing recourse across demographic groups by defining recourse as the distance to a classifier boundary and introducing a regularized objective to minimize disparities in this distance while maintaining accuracy. It develops an explicit, iterative framework for linear and kernelized SVMs, and a model-agnostic, LIME-based approach to approximate recourse for black-box classifiers, including a reweighting scheme to steer decision boundaries. Empirical results on multiple datasets show meaningful reductions in cross-group recourse disparity with minimal loss to predictive performance, across both SVM and model-agnostic settings. This work enables post-decision agency improvements by ensuring more uniform opportunities for individuals to change unfavorable outcomes, applicable to a range of decision-making domains.

Abstract

The rise in machine learning-assisted decision-making has led to concerns about the fairness of the decisions and techniques to mitigate problems of discrimination. If a negative decision is made about an individual (denying a loan, rejecting an application for housing, and so on) justice dictates that we be able to ask how we might change circumstances to get a favorable decision the next time. Moreover, the ability to change circumstances (a better education, improved credentials) should not be limited to only those with access to expensive resources. In other words, \emph{recourse} for negative decisions should be considered a desirable value that can be equalized across (demographically defined) groups. This paper describes how to build models that make accurate predictions while still ensuring that the penalties for a negative outcome do not disadvantage different groups disproportionately. We measure recourse as the distance of an individual from the decision boundary of a classifier. We then introduce a regularized objective to minimize the difference in recourse across groups. We explore linear settings and further extend recourse to non-linear settings as well as model-agnostic settings where the exact distance from boundary cannot be calculated. Our results show that we can successfully decrease the unfairness in recourse while maintaining classifier performance.

Paper Structure

This paper contains 16 sections, 23 equations, 5 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Our Work.
  3. Related work
  4. Definitions
  5. Note.
  6. Optimizing for recourse
  7. Optimizing for recourse with an explicit classifier
  8. An iterative procedure
  9. An example.
  10. Model agnostic settings
  11. Algorithm.
  12. Experiments
  13. Experimental settings.
  14. SVM results
  15. Model agnostic results
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Equalizing recourse across groups, from left (vanilla SVM ) to right ($\lambda$ increased).
  • Figure 2: Recourse regularized SVM classification
  • Figure 3: Recourse equalization in black-box classification
  • Figure 5: Linear (upper) and polynomial (lower) kernels' results on recourse and accuracy. Yellow circle is the mean, and the orange line is the median.
  • Figure 6: RandomForest (left), LogisticRegression (middle), AdaBoost (right) classifier's results on recourse and accuracy. Yellow circle is the mean and the orange line is the median.