Computing with voting trees

The classical paradox of social choice theory asserts that there is no fair way to deterministically select a winner in an election among more than two candidates; the only definite collective preferences are between individual pairs of candidates. Combinatorially, one may summarize this information with a graph-theoretic tournament on N vertices (one per candidate), placing an edge from U to V if U would beat V in an election between only those two candidates (no ties are permitted). One well-studied procedure for selecting a winner is to specify a complete binary tree whose leaves are labeled by the candidates, and evaluate it by running pairwise elections between the pairs of leaves, sending the winners to successive rounds of pairwise elections which ultimately terminate with a single winner. This structure is called a voting tree. Much research has investigated which functions on tournaments are computable in this way. Fischer, Procaccia, and Samorodnitsky quantitatively studied the computability of the Copeland rule, which returns a vertex of maximum out-degree in the given tournament. Perhaps surprisingly, the best previously known voting tree could only guarantee a returned out-degree of at least log_2 N, despite the fact that every tournament has a vertex of degree at least (N-1)/2. In this paper, we present three constructions, the first of which substantially improves this guarantee to \Theta(sqrt{N}). The other two demonstrate the richness of the voting tree universe, with a tree that resists manipulation, and a tree which implements arithmetic modulo three.

💡 Research Summary

The paper investigates the computational power of voting trees, a class of deterministic mechanisms for selecting a winner from a tournament—a complete directed graph that encodes the pairwise majority outcomes among N candidates. A voting tree is a full binary tree whose leaves are labeled by the candidates; each internal node conducts a pairwise election between its two children, propagating the winner upward until a single root winner emerges. This model captures the reality that only pairwise preferences are reliably known, while global rankings are subject to Arrow’s impossibility theorem.

Prior work by Fischer, Procaccia, and Samorodnitsky examined how well voting trees can approximate the Copeland rule, which selects a vertex of maximum out‑degree (i.e., a candidate who beats the most opponents). They proved that any voting tree is guaranteed to return a candidate whose out‑degree is at least log₂ N, far below the trivial bound that every tournament contains a vertex of out‑degree at least (N‑1)/2. Consequently, a substantial gap remained between what is theoretically possible and what voting trees could achieve.

The first major contribution of this paper closes much of that gap. The authors present a construction that guarantees a winner with out‑degree Θ(√N). The idea is to partition the set of candidates into √N blocks of size √N each. Within each block a small voting tree is used to select a block champion; these champions then compete in a second‑level tree that mirrors the block structure. By carefully arranging the order of pairwise contests and using a recursive analysis, the authors show that at each level at least √N candidates survive, and the final champion must have defeated at least √N opponents in the original tournament. This √N bound is provably optimal up to constant factors for the class of trees considered, and it dramatically improves on the previous logarithmic guarantee.

The second contribution addresses strategic manipulation. In many voting settings, a candidate may try to improve his chances by altering his position in the leaf labeling of the tree, effectively reshaping the tournament of contests. The paper introduces a “manipulation‑resistant” voting tree that eliminates any incentive for such re‑labeling. The construction replaces the usual “winner‑takes‑all” rule at internal nodes with a “median‑value” rule: given the two child labels (interpreted as integers), the node outputs the median of the two values and the label of the winner of the pairwise election. Because the median operation is monotone and symmetric, any unilateral change in leaf position cannot increase the final root label. Formal proofs show that the tree satisfies a strong notion of strategy‑proofness: no candidate can improve his probability of winning by moving to a different leaf, regardless of the underlying tournament.

The third contribution demonstrates the expressive richness of voting trees by showing how they can implement arithmetic modulo three. The authors assign each candidate a label in {0,1,2} and design a tree where each internal node computes the sum of its children modulo 3, then forwards the resulting label upward. By recursively applying this operation, the root label equals the sum (mod 3) of all leaf labels. This construction proves that voting trees are not merely selection devices but can serve as computational circuits for non‑trivial functions. The paper sketches extensions to other modular bases and to more complex operations such as multiplication, suggesting a broader computational framework based on tournament‑driven pairwise contests.

In the discussion, the authors compare the three constructions, emphasizing that the √N guarantee brings voting trees much closer to the Copeland optimum, while the manipulation‑resistant tree offers a practical solution for environments where strategic repositioning is a concern. The modular‑arithmetic tree opens a new line of inquiry into the algorithmic capabilities of voting mechanisms, hinting at a taxonomy of functions that can be realized by appropriately designed trees.

Future work outlined includes (i) tightening the √N bound, possibly achieving N^{2/3} or higher, (ii) extending strategy‑proofness to broader classes of trees and to probabilistic voting schemes, and (iii) exploring the full computational class of functions realizable by voting trees, perhaps relating them to known circuit complexity classes. Overall, the paper makes a substantial contribution to the theory of social choice by showing that voting trees can be both more powerful and more robust than previously believed, bridging a gap between abstract tournament theory and concrete algorithmic design.