From Ideals to Institutions: Transaction Costs, Centralization, and Public Choice in Decentralized Autonomous Organizations

Abstract

Decentralized Autonomous Organizations (DAOs) are often portrayed as exemplars of disintermediated, democratic governance, but research suggests this is a myth. We build on that research using data from Boardroom through 2025. We introduce a novel measure of coordination intensity by mapping each DAO's mission to O*NET task importance scores. We find no direct relationship between task complexity and baseline governance centralization. Differences-in-differences estimators are used that leverage the release of advanced Large Language Models (LLMs) in late 2022 as a natural experiment. The introduction of information-processing technology led to a 31\% decrease in direct participation in high-complexity governance decisions. We argue this reflects the emergence of AI-assisted ``governance specialists'' who lower the cost of becoming trusted delegates, accelerating a shift from direct to representative democracy within DAOs. Rather than democratizing participation, LLMs appear to have facilitated functional specialization and delegation.

Keywords: DAO governance, decentralization, transaction costs, public choice theory, polycentric institutions, blockchain, institutional design. JEL Codes: D72, DAOs

JEL Classification: D72, D85, H11, L14, O33, P48