The price of fairness for a small number of indivisible items

TL;DR

This paper investigates the cost of fairness in allocating indivisible goods, showing that the price of fairness grows sublinearly with the number of items relative to agents, and provides exact asymptotics and an efficient formulation.

Contribution

It introduces a new analysis of the price of fairness considering both agents and items, with exact asymptotics and an integer programming approach.

Findings

Price of fairness grows sublinearly when items are not much more than agents.

Exact asymptotics are derived for the case where number of items equals number of agents.

An efficient integer programming formulation for the problem is provided.

Abstract

Incorporating fairness criteria in optimization problems comes at a certain cost, which is measured by the so-called price of fairness. Here we consider the allocation of indivisible goods. For envy-freeness as fairness criterion it is known from literature that the price of fairness can increase linearly in terms of the number of agents. For the constructive lower bound a quadratic number of items was used. In practice this might be inadequately large. So we introduce the price of fairness in terms of both the number of agents and items, i.e., key parameters which generally may be considered as common and available knowledge. It turned out that the price of fairness increases sublinear if the number of items is not too much larger than the number of agents. For the special case of coincide of both counts exact asymptotics could be determined. Additionally an efficient integer programming formulation is given.