We establish a compatibility between fairness and efficiency, captured via Nash Social Welfare (NSW), under the broad class of subadditive valuations. We prove that, for subadditive valuations, there always exists a partial allocation that is envy-free up to the removal of any good (EFx) and has NSW at least half of the optimal; here, optimality is considered across all allocations, fair or otherwise. We also prove, for subadditive valuations, the universal existence of complete allocations that are envy-free up to one good (EF1) and also achieve a factor $1/2$ approximation to the optimal NSW. Our EF1 result resolves an open question posed by Garg, Husic, Li, Végh, and Vondrák (STOC 2023). In addition, we develop a polynomial-time algorithm which, given an arbitrary allocation $\widetilde{A}$ as input, returns an EF1 allocation with NSW at least $\frac{1}{e^{2/e}}\approx \frac{1}{2.08}$ times that of $\widetilde{A}$. Therefore, our results imply that the EF1 criterion can be attained simultaneously with a constant-factor approximation to optimal NSW in polynomial time (with demand queries), for subadditive valuations. The previously best-known approximation factor for optimal NSW, under EF1 and among $n$ agents, was $O(n)$ -- we improve this bound to $O(1)$. It is known that EF1 and exact Pareto efficiency (PO) are incompatible with subadditive valuations. Complementary to this negative result, the current work shows that we regain compatibility by just considering a factor $1/2$ approximation: EF1 can be achieved in conjunction with $\frac{1}{2}$-PO under subadditive valuations. As such, our results serve as a general tool that can be used as a black box to convert any efficient outcome into a fair one, with only a marginal decrease in efficiency.
翻译:我们在次可加估值这一广泛类别下,建立了公平性与效率(通过纳什社会福利(NSW)衡量)之间的兼容性。我们证明,对于次可加估值,总存在一个部分分配,该分配在移除任意物品后是无嫉妒的(EFx),并且其NSW至少达到最优值的一半;此处的最优性是在所有分配(无论公平与否)中考虑的。我们还证明,对于次可加估值,普遍存在完整的分配,这些分配在移除一个物品后是无嫉妒的(EF1),并且也能达到最优NSW的1/2近似因子。我们的EF1结果解决了Garg、Husic、Li、Végh和Vondrák(STOC 2023)提出的一个开放性问题。此外,我们开发了一种多项式时间算法,该算法以任意分配$\\widetilde{A}$作为输入,返回一个EF1分配,其NSW至少是$\\widetilde{A}$的$\\frac{1}{e^{2/e}}\\approx \\frac{1}{2.08}$倍。因此,我们的结果表明,对于次可加估值,EF1准则可以与最优NSW的常数因子近似在多时间内(通过需求查询)同时实现。先前在EF1条件下、在$n$个智能体间,最优NSW的最佳已知近似因子是$O(n)$——我们将此界限改进为$O(1)$。已知EF1与精确帕累托效率(PO)在次可加估值下是不兼容的。作为这一否定结果的补充,当前工作表明,我们仅通过考虑1/2近似因子即可重获兼容性:在次可加估值下,EF1可以与$\\frac{1}{2}$-PO同时实现。因此,我们的结果作为一个通用工具,可以作为一个黑盒使用,将任何高效结果转换为公平结果,而效率仅边际下降。