Revisiting Social Welfare in Bandits: UCB is (Nearly) All You Need

TL;DR

This paper shows that a simple UCB algorithm, with an initial exploration phase, effectively minimizes fairness-aware Nash regret in stochastic bandits, extending to a broad class of fairness metrics with near-optimal guarantees.

Contribution

It demonstrates that a standard UCB algorithm, combined with initial exploration, suffices for near-optimal Nash regret, removing the need for complex, assumption-heavy algorithms.

Findings

UCB with exploration achieves near-optimal Nash regret.

The approach extends to sub-Gaussian rewards.

The method generalizes to p-mean regret with strong guarantees.

Abstract

Regret in stochastic multi-armed bandits traditionally measures the difference between the highest reward and either the arithmetic mean of accumulated rewards or the final reward. These conventional metrics often fail to address fairness among agents receiving rewards, particularly in settings where rewards are distributed across a population, such as patients in clinical trials. To address this, a recent body of work has introduced Nash regret, which evaluates performance via the geometric mean of accumulated rewards, aligning with the Nash social welfare function known for satisfying fairness axioms. To minimize Nash regret, existing approaches require specialized algorithm designs and strong assumptions, such as multiplicative concentration inequalities and bounded, non-negative rewards, making them unsuitable for even Gaussian reward distributions. We demonstrate that an initial uniform exploration phase followed by a standard Upper Confidence Bound (UCB) algorithm achieves near-optimal Nash regret, while relying only on additive Hoeffding bounds, and naturally extending to sub-Gaussian rewards. Furthermore, we generalize the algorithm to a broad class of fairness metrics called the $p$-mean regret, proving (nearly) optimal regret bounds uniformly across all $p$ values. This is in contrast to prior work, which made extremely restrictive assumptions on the bandit instances and even then achieved suboptimal regret bounds.