Implementability of Honest Multi-Agent Sequential Decision-Making with Dynamic Population

• Tác giả: Tao Zhang, Quanyan Zhu
• Ngôn ngữ: eng
• Ký hiệu phân loại: 003.56 Decision theory
• Thông tin xuất bản: 2020
• Bộ sưu tập: Metadata
• ID: 164049

Tóm tắt

We study the design of decision-making mechanism for resource allocations over a multi-agent system in a dynamic environment. Agents' privately observed preference over resources evolves over time and the population is dynamic due to the adoption of stopping rules. The proposed model designs the rules of encounter for agents participating in the dynamic mechanism by specifying an allocation rule and three payment rules to elicit agents' coupled decision makings of honest preference reporting and optimal stopping over multiple periods. The mechanism provides a special posted-price payment rule that depends only on each agent's realized stopping time to directly influence the population dynamics. This letter focuses on the theoretical implementability of the rules in perfect Bayesian Nash equilibrium and characterizes necessary and sufficient conditions to guarantee agents' honest equilibrium behaviors over periods. We provide the design principles to construct the payments in terms of the allocation rules and identify the restrictions of the designer's ability to influence the population dynamics. The established conditions make the designer's problem of finding multiple rules to determine an optimal allocation rule.