Foundations of Mechanism Design for Decentralized Finance
October 21, 2026 (GHC 6501)

Decentralized Finance (DeFi) is currently the largest application of blockchains. However, because today's DeFi mechanisms were not designed to be incentive compatible, participants are incentivized to deviate, and the system's equilibrium behavior is poorly understood. In particular, front-running and back-running attacks, collectively known as Miner Extractable Value (MEV), are rampant. These attacks not only harm ordinary users by driving up prices, but also push the ecosystem toward centralization: today, more than 85% of blocks are produced by the top two block builders.

In this talk, I will propose a new scientific foundation for DeFi mechanism design. We focus on Automated Market Makers (AMMs), arguably the most popular DeFi application today, in which users trade one token $X$ for another token $Y$ through a smart contract on the blockchain. I will present a range of feasibility and infeasibility results and reveal interesting connections to classical algorithmic game theory.

On the one hand, we prove strong impossibility results when the underlying consensus protocol does not prevent censorship. These results provide a game-theoretic justification for the ongoing efforts of many blockchains to upgrade their consensus protocols to support censorship resilience. On the other hand, we present two feasibility results, under different assumptions, that achieve "dream" properties: incentive compatibility for both ordinary users and block builders, and (weak) social welfare optimality. Notably, one of these mechanisms is simple and efficiently computable, making it a strong candidate for practical adoption.