This is the first of three chapters that analyze threshold-based fairness criteria, which are designed to combine a purely utilitarian metric with a maximin or leximax fairness criterion. This chapter focuses on a utility threshold criterion that applies a maximin criterion until the utility cost crosses a threshold, at which point it begins to apply a utilitarian criterion. The threshold is user-specified by a parameter \(\Delta \) , where larger values of \(\Delta \) correspond to greater fairness. The chapter presents a mixed integer programming model of the resulting optimization problem and a validity proof that appears in the literature. It shows that the solution subject to a budget constraint is either purely utilitarian or purely maximin, depending on a computable value of \(\Delta \) . Additionally, it shows that when there are also stakeholder utility bounds, at most one stakeholder’s utility lies strictly between the smallest utility and that stakeholder’s upper bound. A similar property is proved for hierarchical distributions.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Utility Threshold Criterion

  • Özgün Elçi,
  • John Hooker,
  • Peter Zhang

摘要

This is the first of three chapters that analyze threshold-based fairness criteria, which are designed to combine a purely utilitarian metric with a maximin or leximax fairness criterion. This chapter focuses on a utility threshold criterion that applies a maximin criterion until the utility cost crosses a threshold, at which point it begins to apply a utilitarian criterion. The threshold is user-specified by a parameter \(\Delta \) , where larger values of \(\Delta \) correspond to greater fairness. The chapter presents a mixed integer programming model of the resulting optimization problem and a validity proof that appears in the literature. It shows that the solution subject to a budget constraint is either purely utilitarian or purely maximin, depending on a computable value of \(\Delta \) . Additionally, it shows that when there are also stakeholder utility bounds, at most one stakeholder’s utility lies strictly between the smallest utility and that stakeholder’s upper bound. A similar property is proved for hierarchical distributions.