Distortion of Metric Voting with Bounded Randomness

We study the design of voting rules in the metric distortion framework. It is known that any deterministic rule suffers distortion of at least $3$, and that randomized rules can achieve distortion strictly less than $3$, often at the cost of reduced transparency and interpretability. In this work, we explore the trade-off between these paradigms by asking whether it is possible to break the distortion barrier of $3$ using only “bounded” randomness. We answer in the affirmative by presenting a voting rule that (1) achieves distortion of at most $3 - \varepsilon$ for some absolute constant $\varepsilon > 0$, and (2) selects a winner uniformly at random from a deterministically identified list of constant size. Our analysis builds on new structural results for the distortion and approximation of Maximal Lotteries and Stable Lotteries.

💡 Research Summary

The paper investigates voting rule design within the metric distortion framework, focusing on the trade‑off between deterministic transparency and the efficiency gains of randomization. It is well‑known that any deterministic rule must incur a distortion of at least 3, while fully randomized rules can achieve distortion strictly below 3 (e.g., 3 − ε for some constant ε > 0). However, unrestricted randomization harms interpretability and practical acceptance. The authors ask whether one can break the “3‑barrier” using only “bounded” randomness—that is, by randomizing uniformly over a constant‑size list of candidates.

The main contribution is an affirmative answer. The authors present a voting rule that (1) guarantees distortion at most 3 − ε for an absolute constant ε, and (2) selects a winner uniformly at random from a deterministically identified set of constant size k. The rule is computationally efficient: a polynomial‑time algorithm enumerates multisets of candidates of increasing size and stops when the induced uniform distribution achieves distortion below 3 − ε; because such a multiset of constant size is guaranteed to exist, the enumeration terminates after a constant number of steps.

Technical foundations are built on several structural results:

1. Robustness of Maximal Lotteries (ML). While MLs have expected distortion ≤ 3, the authors prove a stronger worst‑case bound: any candidate in the support of an ML yields distortion ≤ 4 + √17 < 8. This shows that ML‑based rules are robust to the actual random outcome, not just in expectation.

2. Approximate Maximal Lotteries. Directly using ML is impossible under bounded randomness because the support of an ML can be linear in the number of candidates. Inspired by Charikar, Ramakrishnan, and Wang (CRW26), the authors sample a constant number of candidates from an ML and consider the uniform distribution over these samples. They prove that, for any prescribed ε > 0, a constant‑size sample (depending only on ε) yields a distribution that, with positive probability, guarantees distortion ≤ 3 + ε. The proof leverages the biased‑metric framework and careful bounding of approximation losses.

3. Stable Lotteries and Their Approximation. When the underlying biased metric is sufficiently consistent, Stable Lotteries (a generalization of ML) provide even stronger guarantees. The authors extend the sampling technique to Stable Lotteries, showing that a uniform distribution over a constant number of samples approximates the distortion of the exact Stable Lottery without dependence on the number of voters or candidates.

4. Biased Metrics and (α,β)‑Consistency. The analysis works within the biased‑metric model introduced by Chen and Roughgarden (CR22). This model captures the hardest metric instances; proving a bound for all biased metrics automatically yields the same bound for all metrics. The authors also rely on the notion of (α,β)‑consistency, which has been central to all prior sub‑3 distortion results.

The paper’s algorithmic construction proceeds in three parts (Section 7). Part I handles inconsistent biased metrics, Part II handles consistent ones, and Part III mixes the two approaches to obtain a single rule that works for any metric space. The resulting rule randomizes over at most a constant number k of candidates, achieving distortion strictly below 3.

Beyond single‑winner elections, the authors discuss implications for committee selection (selecting k winners). They observe that if each candidate may occupy multiple seats, then for sufficiently large committee size there exists a deterministic committee rule with distortion < 3, mirroring the bounded‑randomness result.

Overall, the paper demonstrates that limiting randomness to a constant‑size uniform lottery does not preclude breaking the deterministic distortion barrier. By combining structural insights about Maximal and Stable Lotteries with biased‑metric analysis and sampling concentration (DKW inequality), the authors provide both theoretical guarantees and a practical, transparent voting mechanism that improves upon the classic deterministic bound while preserving interpretability.