Abstract:The two standard fairness notions in the resource allocation literature are proportionality and envy-freeness. If there are n agents competing for the available resources, then proportionality requires that each agent receives at least a 1/n fraction of their total value for the set of resources. On the other hand, envy-freeness requires that each agent weakly prefers the resources allocated to them over those allocated to any other agent. Each of these notions has its own benefits, but it is well known that neither one of the two is always achievable when the resources being allocated are indivisible. As a result, a lot of work has focused on satisfying fairness notions that relax either proportionality or envy-freeness. In this paper, we focus on MXS (a relaxation of proportionality) and EFL (a relaxation of envy-freeness). Each of these notions was previously shown to be achievable on its own [Barman et al.,2018, Caragiannis et al., 2023], and our main result is an algorithm that computes allocations that simultaneously satisfy both, combining the benefits of approximate proportionality and approximate envy-freeness. In fact, we prove this for any instance involving agents with valuation functions that are restricted MMS-feasible, which are more general than additive valuations. Also, since every EFL allocation directly satisfies other well-studied fairness notions like EF1, 1/2-EFX, 1/2-GMMS, and 2/3-PMMS, and every MXS allocation satisfies 4/7-MMS, the allocations returned by our algorithm simultaneously satisfy a wide variety of fairness notions and are, therefore, universally fair [Amanatidis et al., 2020].
What problem does this paper attempt to address?
### The problems the paper attempts to solve
This paper aims to solve the fairness problem in resource allocation. In particular, when resources are indivisible, how to simultaneously meet two relaxed fairness criteria: MXS (approximate proportional fairness) and EFL (approximate envy - freeness). Specifically:
1. **Proportionality**:
- When there are \(n\) agents competing for resources, each agent should receive a resource bundle that is at least \(\frac{1}{n}\) of the total value of all resources. Expressed in a formula:
\[
v_i(X_i)\geq\frac{1}{n}\sum_{g\in M}v_i(g)
\]
where \(v_i\) is the valuation function of agent \(i\), and \(X_i\) is the resource bundle allocated to agent \(i\).
2. **Envy - freeness**:
- Each agent's preference for the resource bundle allocated to it is no lower than that of any other agent's resource bundle. Expressed in a formula:
\[
v_i(X_i)\geq v_i(X_j),\quad\forall i,j\in N
\]
However, when resources are indivisible, these two fairness criteria usually cannot be achieved simultaneously. Therefore, researchers have proposed two relaxed criteria:
- **MXS (Max - Min Share Fairness)**:
- This is a relaxed form of proportional fairness, ensuring that each agent receives a resource bundle that is at least a certain proportion of its max - min share. Expressed in a formula:
\[
v_i(X_i)\geq\alpha\cdot\text{MMS}_i
\]
where \(\alpha\) is a constant, and \(\text{MMS}_i\) is the max - min share of agent \(i\).
- **EFL (Envy - Free up to the Second - Best Item)**:
- This is a relaxed form of envy - freeness, ensuring that each agent will not envy other agents' resource bundles, unless after removing one of the other agent's second - best items, it is still not better than its own resource bundle. Expressed in a formula:
\[
|X_j| > 1\Rightarrow v_i(X_i)\geq v_i(X_j\setminus\{g\}),\quad\forall g\in X_j
\]
### Main contributions of the research
The main contribution of this paper is to propose an algorithm that can calculate a resource allocation scheme that simultaneously satisfies MXS and EFL. This not only combines the advantages of approximate proportional fairness and approximate envy - freeness, but also applies to instances with restricted MMS - feasible valuation functions. In addition, since EFL allocation directly satisfies other known fairness criteria (such as EF1, 1/2 - EFX, 1/2 - GMMS, and 2/3 - PMMS), and MXS allocation satisfies 4/7 - MMS, the allocation scheme returned by this algorithm can simultaneously satisfy multiple fairness criteria, thus achieving "universal fairness".
### Summary
This paper solves the challenging problem of simultaneously satisfying MXS and EFL in the allocation of indivisible resources by introducing a new algorithm, and proves the effectiveness and wide applicability of this algorithm.