We will host an international workshop “2024-25 Rising Stars in Market Design” presented by young researchers on the 2024-2025 academic job market on November 26 (JST).
Event Information
Date & Times: 9:00am – 12:20pm, Tuesday, November 26 (JST) Venue: Zoom online (pre-registration system) Language: English *Please refrain from taking pictures, recording audio and video, or reproducing any content presented during the event. *Please do not share the Zoom joining URL with a third party.
About registration
There is no registration fee, but pre-registration is required. Please sign up through the form below.
Even if you have registered for our workshops or conferences in the past, we kindly ask you to re-register this time as we are using a new meeting URL.
Truthful Aggregation of LLMs with an Application to Online Advertising
10:10-11:10am
Kei Ikegami (New York University)
Bargaining over Leasing Contracts amid Shifting Power Balance
11:20am-12:20pm
Lea Nagel (Stanford University)
As-If Dominant Strategy Mechanisms
*All times are Japan Standard Time.
Speaker
Ermis Soumalias (University of Zurich)
Title: Truthful Aggregation of LLMs with an Application to Online Advertising
Abstract:The next frontier of online advertising is revenue generation from LLM-generated content. We consider a setting where advertisers aim to influence the responses of an LLM to align with their interests, while platforms seek to maximize advertiser value and ensure user satisfaction. The challenge is that advertisers’ preferences generallyconflictwiththoseoftheuser,andadvertisersmaymisreporttheir preferences. To address this, we introduce MOSAIC, an auction mechanism that ensures that truthful reporting is a dominant strategy for advertisers and that aligns the utility of each advertiser with their contribution to social welfare. Importantly, the mechanism operates without LLM fine-tuning or access to model weights and provably converges to the output of the optimally fine-tuned LLM as computational resources increase. Additionally, it can incorporate contextual information about advertisers, which significantly improves social welfare. Through experiments with a publicly available LLM, we show that MOSAIC leads to high advertiser value and platform revenue with low computational overhead.While our motivating application is online advertising, our mechanism can be applied in any setting with monetary transfers, making it a general-purpose solution for truthfully aggregating the preferences of self-interested agents over LLM-generated replies.
Kei Ikegami (New York University)
Title: Bargaining over Leasing Contracts amid Shifting Power Balance
Abstract:This paper presents a model where contract renewals and terms are negotiated based on past performance, which influences renewal through two channels: altering the bargaining environment and shifting the power balance between parties. Applied to shopping mall leasing contracts, my analysis reveals that fluctuations in the balance of bargaining powers alter the mall’s share of surplus by10%. Controlling for bargaining power enables a clearer identification of contract term determinants. Notably, the mall managers’ risk preferences affect contract selection, with the potential to double rental income by changing the managers under the same power balance with the tenants. JEL Classification Codes: C71; C78; L81; R32; R33. Keywords: Contracting; Nash Bargaining; Renegotiation; Shopping Center; Tenancy; Tenant.
Lea Nagel (Stanford University)
Title: As-If Dominant Strategy Mechanisms
Abstract:Weshowthatachievingdominantstrategyincentivecompatibilityoftenrequirestochooseamechanismwhichseverelylimitswhatagentscanobserve about others’ previous moves. However, experiments and theoretical arguments suggest increasing the transparency of a mechanism’s extensive form can improve reliability of its predictions—even if it breaks the dominant strategy property. To help resolve this dilemma, we defineas-if dominant strategy mechanisms:(i) Eachagenthasatleastonestrategythatbecomesdominantiftheotherswere restrictedtobehaveasifthemechanismwasstatic,and(ii)allcombinationsof suchstrategiesareex-postequilibria.Thesemechanismslooklikeadominant strategy one to cognitively limited agents who neglect others may condition their behaviorinsophisticatedways,andcanhelpthemavoiddominatedbehaviors. Moreover, they ensure sophisticated agents never have an incentive to deviate. Ourframeworkrationalizestheauctionformatchosenbyprominentonline platforms,suchaseBay.Italsoprovidesaunifiedexplanationforexperimental evidenceinvarioussettings.Further,weprovidesufficientconditionsforas-if dominant strategy mechanisms to also be weak dominance solvable.
Contact info
The University of Tokyo Market Design Center（UTMD） Graduate School of Economics, the University of Tokyo E-mail: market-design[at]e.u-tokyo.ac.jp Phone: (+81)3-5841-3441
