Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

End-to-End Optimization and Learning of Fair Court Schedules

My H Dinh, James Kotary, Lauryn P. Gouldin, William Yeoh, Ferdinando Fioretto

TL;DR

A joint optimization and learning framework that combines machine learning models trained end-to-end with efficient matching algorithms is proposed that aims to produce court scheduling schedules that optimize a principled measure of fairness, balancing the availability and preferences of all parties.

Abstract

Criminal courts across the United States handle millions of cases every year, and the scheduling of those cases must accommodate a diverse set of constraints, including the preferences and availability of courts, prosecutors, and defense teams. When criminal court schedules are formed, defendants' scheduling preferences often take the least priority, although defendants may face significant consequences (including arrest or detention) for missed court dates. Additionally, studies indicate that defendants' nonappearances impose costs on the courts and other system stakeholders. To address these issues, courts and commentators have begun to recognize that pretrial outcomes for defendants and for the system would be improved with greater attention to court processes, including \emph{court scheduling practices}. There is thus a need for fair criminal court pretrial scheduling systems that account for defendants' preferences and availability, but the collection of such data poses logistical challenges. Furthermore, optimizing schedules fairly across various parties' preferences is a complex optimization problem, even when such data is available. In an effort to construct such a fair scheduling system under data uncertainty, this paper proposes a joint optimization and learning framework that combines machine learning models trained end-to-end with efficient matching algorithms. This framework aims to produce court scheduling schedules that optimize a principled measure of fairness, balancing the availability and preferences of all parties.

End-to-End Optimization and Learning of Fair Court Schedules

TL;DR

A joint optimization and learning framework that combines machine learning models trained end-to-end with efficient matching algorithms is proposed that aims to produce court scheduling schedules that optimize a principled measure of fairness, balancing the availability and preferences of all parties.

Abstract

Criminal courts across the United States handle millions of cases every year, and the scheduling of those cases must accommodate a diverse set of constraints, including the preferences and availability of courts, prosecutors, and defense teams. When criminal court schedules are formed, defendants' scheduling preferences often take the least priority, although defendants may face significant consequences (including arrest or detention) for missed court dates. Additionally, studies indicate that defendants' nonappearances impose costs on the courts and other system stakeholders. To address these issues, courts and commentators have begun to recognize that pretrial outcomes for defendants and for the system would be improved with greater attention to court processes, including \emph{court scheduling practices}. There is thus a need for fair criminal court pretrial scheduling systems that account for defendants' preferences and availability, but the collection of such data poses logistical challenges. Furthermore, optimizing schedules fairly across various parties' preferences is a complex optimization problem, even when such data is available. In an effort to construct such a fair scheduling system under data uncertainty, this paper proposes a joint optimization and learning framework that combines machine learning models trained end-to-end with efficient matching algorithms. This framework aims to produce court scheduling schedules that optimize a principled measure of fairness, balancing the availability and preferences of all parties.

Paper Structure

This paper contains 28 sections, 1 theorem, 16 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Court scheduling.
  4. Advances in Machine Learning and Optimization for scheduling.
  5. Motivations and Problem Setting
  6. Learning Objective
  7. Preliminaries: Fair OWA Aggregation
  8. Fair Optimization of Court Schedules
  9. Group Fairness
  10. Complexity of the Optimization Models
  11. Optimization and Learning for Fair Court Schedules
  12. End-to-End Trainable Scheduling Model
  13. Differentiable Matching Layer
  14. OWA as a Loss Function
  15. Experimental Settings
  16. ...and 13 more sections

Key Result

Proposition 1

The matching layer model:matching_layer has complexity $\mathcal{O}(n^3)$.

Figures (8)

  • Figure 1: Proposed end-to-end framework for learning to schedule. Given candidates' socioeconomic and demographic identifiers, a neural network is trained to predict their preference score for each time slot. A differentiable surrogate model uses a predicted score to attain assignments and decision quality loss.
  • Figure 2: Court scheduling example
  • Figure 3: Causal relationship depicts factors affecting individual's court schedule preference. Blue arrows indicate indirect relationships, while green arrows indicate direct relationships.
  • Figure 4: OWA utility regret \ref{['eq:regret']} in court scheduling across different fairness levels: individual (first subplot) vs. group (last three subplot). Protected group attributes are employment status, transportation accessibility, and work hours.
  • Figure 5: Benchmarking OWA Regret (in percentage) across different training data sizes.
  • ...and 3 more figures

Theorems & Definitions (2)

  • Proposition 1
  • proof