The computational complexity of winner determination is a classical and important problem in computational social choice. Previous work based on worst-case analysis has established NP-hardness of winner determination for some classic voting rules, such as Kemeny, Dodgson, and Young. In this paper, we revisit the classical problem of winner determination through the lens of semi-random analysis, which is a worst average-case analysis where the preferences are generated from a distribution chosen by the adversary. Under a natural class of semi-random models that are inspired by recommender systems, we prove that winner determination remains hard for Dodgson, Young, and some multi-winner rules such as the Chamberlin-Courant rule and the Monroe rule. Under another natural class of semi-random models that are extensions of the Impartial Culture, we show that winner determination is hard for Kemeny, but is easy for Dodgson. This illustrates an interesting separation between Kemeny and Dodgson.
翻译:确定赢家的计算复杂性是计算社会选择中的一个传统而重要的问题。 以往基于最坏案例分析的工作已经确立了某些经典投票规则,如凯梅尼、多格森和杨等经典投票规则的确定赢家的硬度。 在本文中,我们通过半随机分析的透镜重新审视了确定赢家的典型问题,半随机分析是一种最差的平均案例分析,其偏好来自对手选择的分布。在一个由推荐人系统启发的半随机模型的自然类别下,我们证明,对Dodgson、Young和一些多赢家规则,如Camberlin-Courant规则和Monroe规则来说,赢家的确定仍然是困难的。在另一个自然的半随机模型类别中,我们展示了对Kemeny来说很难确定赢家,但对Dodgson来说是容易的。 这显示了Kemeny和Dodgson之间有趣的区分。