Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scarce Resource Allocations That Rely On Machine Learning Should Be Randomized

Shomik Jain, Kathleen Creel, Ashia Wilson

TL;DR

The paper tackles fairness in scarce resource allocations governed by ML where deterministic decisions may perpetuate inequities. It proposes stochastic procedures grounded in Broome's theory of claims, chiefly weighted lotteries (BF) and partial BF variants, to respect individual claims under uncertainty and across multiple decision-makers. The authors develop a formal framework for known and unknown claims, and demonstrate that small amounts of randomization can substantially reduce systemic exclusion rate with limited utility loss, via simulations and two real-data examples (Swiss unemployment and Census income). They provide practical guidelines, discuss scope and ethics, and supply code to implement these methods.

Abstract

Contrary to traditional deterministic notions of algorithmic fairness, this paper argues that fairly allocating scarce resources using machine learning often requires randomness. We address why, when, and how to randomize by proposing stochastic procedures that more adequately account for all of the claims that individuals have to allocations of social goods or opportunities.

Scarce Resource Allocations That Rely On Machine Learning Should Be Randomized

TL;DR

The paper tackles fairness in scarce resource allocations governed by ML where deterministic decisions may perpetuate inequities. It proposes stochastic procedures grounded in Broome's theory of claims, chiefly weighted lotteries (BF) and partial BF variants, to respect individual claims under uncertainty and across multiple decision-makers. The authors develop a formal framework for known and unknown claims, and demonstrate that small amounts of randomization can substantially reduce systemic exclusion rate with limited utility loss, via simulations and two real-data examples (Swiss unemployment and Census income). They provide practical guidelines, discuss scope and ethics, and supply code to implement these methods.

Abstract

Contrary to traditional deterministic notions of algorithmic fairness, this paper argues that fairly allocating scarce resources using machine learning often requires randomness. We address why, when, and how to randomize by proposing stochastic procedures that more adequately account for all of the claims that individuals have to allocations of social goods or opportunities.
Paper Structure (17 sections, 14 figures, 6 tables, 3 algorithms)

This paper contains 17 sections, 14 figures, 6 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Why and When To Randomize Allocations
  4. Fairness and Individual Claims
  5. Multi-Shot Contexts: Systemic Denial of Claims
  6. Inherent Uncertainties in Predicting Claims
  7. How To Randomize Allocations
  8. When Claims Are Known
  9. Systemic Harms
  10. Utility and Randomization
  11. When Claims Are Uncertain
  12. Discussion
  13. Appendix
  14. When Claims Are Known Experiments
  15. When Claims Are Unknown Experiments
  16. ...and 2 more sections

Figures (14)

  • Figure 1: Normal Distribution of Claims
  • Figure 2: Inverse Normal Distribution of Claims
  • Figure 3: Pareto Distribution of Claims
  • Figure 4: Inverse Pareto Distribution of Claims
  • Figure 5: Uniform Distribution of Claims
  • ...and 9 more figures

Theorems & Definitions (10)

  • Definition 4.1
  • Definition 4.2
  • Example 4.3: BF Lottery
  • Definition 4.4
  • Example 4.5: Systemic Exclusion
  • Example 4.6: Patterned Inequality
  • Definition 4.7
  • Definition 4.8
  • Definition 4.9
  • Example 4.10: Partial BF Lottery