论文标题
信号的力量及其与无政府状态价格的内在联系
The Power of Signaling and its Intrinsic Connection to the Price of Anarchy
论文作者
论文摘要
游戏理论的一个基本课程是,战略行为通常会使均衡结果效率低下。信息设计的最新文献(又称信号或说服力)旨在通过向玩家提供精心调整的信息来改善均衡,以影响他们的行动。以前的大多数研究都集中在设计最佳信号方案的问题上。这项工作与以前的研究相偏离了描述性问题,并希望定量地表征信号传导的力量(POS),即信号设计师可以改善其平衡结果的目标。我们考虑具有越来越多的功率的四个信号计划:完整信息,最佳的公共信号,最佳的私人信号和最佳的前私人信号传导。我们的主要结果是,在广泛的成本最小化和最大最大化游戏中,每个信号方案都对其直接的前身具有的额外功率进行了干净而紧密的表征,其中:(1)所有玩家都将非负成本函数最小化或最大化非负收益功能; (2)信号设计师(自然)优化了玩家实用程序的总和。我们证明,信号传导的附加功率 - 定义为上述两个连续信号方案的平衡目标之间的最差比率 - 精确地限制在相应游戏的无政府状态价格的良好概念上。此外,我们表明所有这些界限都很紧。
A basic lesson from game theory is that strategic behavior often renders the equilibrium outcome inefficient. The recent literature of information design -- a.k.a. signaling or persuasion -- looks to improve equilibria by providing carefully-tuned information to players in order to influence their actions. Most previous studies have focused on the question of designing optimal signaling schemes. This work departs from previous research by considering a descriptive question and looks to quantitatively characterize the power of signaling (PoS), i.e., how much a signaling designer can improve her objective of the equilibrium outcome. We consider four signaling schemes with increasing power: full information, optimal public signaling, optimal private signaling and optimal ex-ante private signaling. Our main result is a clean and tight characterization of the additional power each signaling scheme has over its immediate predecessor above in the broad classes of cost-minimization and payoff-maximization games where: (1) all players minimize non-negative cost functions or maximize non-negative payoff functions; (2) the signaling designer (naturally) optimizes the sum of players' utilities. We prove that the additional power of signaling -- defined as the worst-case ratio between the equilibrium objectives of two consecutive signaling schemes in the above list -- is bounded precisely by the well-studied notion of the price of anarchy of the corresponding games. Moreover, we show that all these bounds are tight.