[Colloquium] Minbiao Han MS Presentation/Jul 18, 2023

[Megan Woodward]

This is an announcement of Minbiao Han's MS Presentation
===============================================
Candidate: Minbiao Han

Date: Tuesday, July 18, 2023

Time:  1:30 pm CST

Remote Location:  https://uchicago.zoom.us/j/97790413594?pwd=NHJSVGU0a3hROGgvQndicnd5SHlKZz09&from=addon

Location: JCL 390

M.S. Paper Title: From Dynamic Pricing to Dynamic Principal-Agent Problems: Going Beyond the No Learning Theorem

Abstract: This paper studies dynamic principal-agent problems, i.e., games in which a principal and an agent repeatedly interact, where the agent’s type is unknown and the agent is non-myopic. A natural question to ask is, can the principal learn the optimal strategy against the unknown agent through these repeated interactions? The No Learning Theorem from dynamic pricing, a special class of dynamic principal-agent problems, would suggest that the principal cannot learn effectively from the agent. In contrast, we demonstrate, for the first time, that in general principal-agent problems, the principal can improve her utility through learning in repeated plays. We show that this dynamic policy continues to be nearly optimal even when allowing for the principal to have a larger static strategy space, specifically if we permit communication between the principal and agent. We also provide an algorithm based on a novel and compact mixed-integer linear program for finding the principal’s optimal dynamic policy. In addition, we develop an algorithm to compute a Markovian policy for the principal that approximates the optimal dynamic policy while allowing for more efficient computation. Through simulations, we examine the efficiency, compared to static policies, and the runtime of the proposed algorithms. Lastly, we apply the generalized principal-agent framework to a specific contract design problem. We show that with the special structural properties of contract design, the optimal dynamic principal policy has a compact representation. Furthermore, in the case where the agent’s type is known, dynamic principal policies enable full surplus extraction from the agent when the interaction with the principal extends over a sufficiently long time horizon.

Advisors: Haifeng Xu

Committee Members: Haifeng Xu, Yuxin Chen, Cong Ma, and Michael Albert





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20230703/81d44937/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: MinbiaoHan_MS_Exam.pdf
Type: application/pdf
Size: 687810 bytes
Desc: MinbiaoHan_MS_Exam.pdf
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20230703/81d44937/attachment-0001.pdf>