A Unified Framework of Multi-Stage Multi-Winner Voting: An Axiomatic Exploration
Shengjie Gong, Lingxiao Huang, Shuangping Huang, Yuyi Wang, Zhiqi Wang, Tao Xiao, Xiang Yan, Chunxue Yang
TL;DR
Multi-winner elections face manipulation and strategic voting; this paper proposes a unified multi-stage framework $\mathcal{R}=(R_1,\ldots,R_t)$ to mitigate manipulation costs by successive shortlists. It develops a formal score-based rule class using parameters $(\beta,\gamma)$ and shows how classic rules such as SNTV, Bloc, Borda, and CC fit into the framework, with STV and Baldwin's rule as multi-stage instances. The authors prove that Solid Coalition is preserved across stages when stage rules satisfy it, but demonstrate that Committee Monotonicity, Candidate Monotonicity, and Consistency generally fail in the multi-stage setting under rational rules. They provide guidance on rule selection, suggesting using SC-preserving stage rules like SNTV and noting that approval-based Thiele rules exhibit stronger multi-stage monotonicity properties. The work lays a theoretical foundation for multi-stage voting design and highlights directions for future axiomatic analyses and manipulation-resistance investigations.
Abstract
Multi-winner voting plays a crucial role in selecting representative committees based on voter preferences. Previous research has predominantly focused on single-stage voting rules, which are susceptible to manipulation during preference collection. In order to mitigate manipulation and increase the cost associated with it, we propose the introduction of multiple stages in the voting procedure, leading to the development of a unified framework of multi-stage multi-winner voting rules. To shed light on this framework of voting methods, we conduct an axiomatic study, establishing provable conditions for achieving desired axioms within our model. Our theoretical findings can serve as a guide for the selection of appropriate multi-stage multi-winner voting rules.
