CAREER: Towards a Robust Theory of Mechanism Design

职业生涯：建立稳健的机构设计理论

基本信息

批准号：
1942583
负责人：
Yang Cai
金额：
$ 60万
依托单位：
Yale University
依托单位国家：
美国
项目类别：
Continuing Grant
财政年份：
2020
资助国家：
美国
起止时间：
2020-01-01 至 2024-12-31
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1942583&HistoricalAwards=false
关键词：
CAREER Towards Robust Theory Mechanism

项目摘要

The ever-growing technology platforms such as sponsored search, online marketplaces, crowdsourcing, and sharing economies are becoming cornerstones of the modern economy. A central problem faced by these online markets and platforms is how to design the right incentive structure, so that the participants, who are only interested in optimizing their own utilities, are motivated to take actions that will help to realize the designer's goals. The classic economic theory of mechanism design is dedicated to addressing this problem. However, it is not yet robust or realistic enough to provide concrete guidelines for practice due to several shortcomings such as the strong reliance on distributional assumptions, the focus on optimal but impractical mechanisms, as well as the lack of attention to the role of information. The goal of this research is to build a robust mechanism design theory by offering new frameworks and solutions to alleviate and resolve these shortcomings. In tackling these weaknesses, this project not only improves the practicality of mechanism design, but also develops insights to answer some of the long-standing theoretical open questions. This project includes an education plan that incorporates course development of both graduate and undergraduate courses as well as training for graduate students and research opportunities for undergraduates. More concretely, this project focuses on the following three research thrusts. (i) Weakening Bayesian assumptions: a common but unrealistic assumption in mechanism design is that the participants' preferences are drawn from a known distribution. The investigator plans to revisit mechanism design under two alternative and realistic distribution access models: (a) sample access to the distribution and (b) max-min robust learning -- given only an approximate distribution, learn a mechanism that performs well under the unknown true distribution. (ii) Understanding the tradeoff between simplicity and optimality: the optimal mechanism is usually too complex to be practical. The investigator will develop a framework to design simple and approximately optimal mechanisms and address open questions in multi-item auctions and two-sided markets. (iii) Information design via an algorithmic lens: unlike traditional mechanism design, which focuses on influencing participants via direct incentives, information design studies how information revelation can shift the participants’ behavior. The investigator aims to understand the design of information structure via an algorithmic lens, focusing on the computational complexity of the optimal information revelation scheme. This work will rely on tools from optimization, learning theory, statistics and machine learning, and statistical physics, and forge new connections between these fields and mechanism design.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

不断增长的技术平台，如赞助搜索、在线市场、众包和共享经济，正在成为现代经济的基石，这些在线市场和平台面临的一个中心问题是如何设计正确的激励结构，使参与者能够参与其中。只对优化自己的效用感兴趣，有动力采取有助于实现设计者目标的行动。机制设计的经典经济理论致力于解决这个问题，但它还不够稳健或现实。为实践提供具体的指导方针缺点，例如强烈依赖分布假设、关注最优但不切实际的机制以及缺乏对信息作用的关注。本研究的目标是通过提供新的框架和解决方案来建立稳健的机制设计理论。为了缓解和解决这些缺点，该项目不仅提高了机制设计的实用性，而且还提出了一些见解来回答一些长期存在的理论开放问题。研究生和本科生课程更具体地说，该项目侧重于以下三个研究重点：（i）弱化贝叶斯假设：机制设计中的一个常见但不切实际的假设是参与者的偏好来自于。研究人员计划在两种替代且现实的分布访问模型下重新审视机制设计：（a）对分布的样本访问和（b）最大-最小鲁棒学习——仅给定近似分布，学习表现良好的机制。在未知的真实分布下。 (ii) 理解简单性和最优性之间的权衡：最优机制通常过于复杂而不实用。研究者将开发一个框架来设计简单且近似最优的机制，并解决多项目拍卖和双边市场中的开放性问题。（iii）通过算法视角进行信息设计：与传统的机制设计侧重于通过直接激励影响参与者不同，信息设计研究信息揭示如何改变参与者的行为，研究者旨在通过算法来理解信息结构的设计。镜头，关注最优信息揭示的计算复杂性这项工作将依赖于优化、学习理论、统计学和机器学习以及统计物理学等领域的工具，并在这些领域和机制设计之间建立新的联系。该奖项是 NSF 的法定使命，并通过使用评估被认为值得支持。基金会的智力价值和更广泛的影响审查标准。

项目成果

期刊论文数量（15）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Multi-Item Mechanisms without Item-Independence: Learnability via Robustness

没有项目独立性的多项目机制：通过鲁棒性实现可学习性

DOI：
10.1145/3391403.3399541
发表时间：
2020
期刊：
21st ACM Conference on Economics and Computation
影响因子：
0
作者：
Brustle, Johannes;Cai, Yang;Daskalakis, Constantinos
通讯作者：
Daskalakis, Constantinos

Is Selling Complete Information (Approximately) Optimal?

出售（大约）完整信息是最佳选择吗？

DOI：
10.1145/3490486.3538304
发表时间：
2022
期刊：
Proceedings of the 23rd ACM Conference on Economics and Computation
影响因子：
0
作者：
Bergemann, Dirk;Cai, Yang;Velegkas, Grigoris;Zhao, Mingfei
通讯作者：
Zhao, Mingfei

How to Sell Information Optimally: An Algorithmic Study

如何最佳地销售信息：算法研究

DOI：
10.4230/lipics.itcs.2021.81
发表时间：
2021
期刊：
Proceedings of the12th Innovations in Theoretical Computer Science Conference
影响因子：
0
作者：
Cai, Yang;Velegkas, Grigoris
通讯作者：
Velegkas, Grigoris

Finite-Time Last-Iterate Convergence for Learning in Multi-Player Games