Agents' Behavior and Interest Rate Model Optimization in DeFi Lending

ABSTRACT Contrasting sharply with traditional money, bond, and bond futures markets, where interest rates emerge organically from participant interactions, DeFi lending platforms employ rule‐based interest rates that are algorithmically set. Thus, the selection of an effective interest rate model (IRM) is paramount for the success of a lending protocol. This paper investigates the modeling of agents' behaviors on lending platforms and proposes a theoretical framework for formulating optimal IRMs. We show that, under perfect information, an optimal control model with a state constraint generates an optimal interest rate policy that has a shape similar to that of popular markets. Furthermore, we formally analyze interest rate policies based on PID controllers, which work efficiently based on fewer assumptions. Using public data of popular markets on the Ethereum blockchain, we analyze agents' behavior, build a realistic simulation environment, and highlight the main tradeoffs in the design of interest rates for decentralized lending platforms.