Six Trends in Decentralized Governance: A Look Ahead to 2025
As the online world becomes increasingly important in our lives, the critical collective decisions we need to make on the internet will also continue to grow. As a result, 2025 is an exciting year for decentralized governance: DAOs will continue to explore new ways to allow anonymous token holders to govern collectively; at the same time, large institutions like Blackrock, State Street, and Vanguard are trying to get their clients more involved in online shareholder voting; AI companies like Anthropic, Meta, and OpenAI are using citizen assemblies to set usage rules for their large language models (LLMs).
Riding this strong momentum, here are some themes I will continue to follow and engage with next year.
1. Websites to Facilitate Proxy Voting
Background: Proxy voting is crucial for online governance, but how to help voters choose their proxies wisely remains a challenge. Research shows that user-friendly proxy voting websites can increase participation by over 20%, but when there are many proxy options to choose from (a good problem to have), how to present this information without further concentrating power becomes crucial. We're still uncertain about the best ways to display proxy information and who should be responsible for this important presentation. 2025 will be a pivotal year for figuring out how to present proxy information and track its impact.
2. AI Assistants to Help with Proxy Voting
Background: Can we use AI agents to help users find proxy voters that align with their views? Perhaps users could engage in a dialogue with an AI agent, answering questions about their values and goals; the AI agent would then review the voting records, manifestos, proposals, and forum discussions of the proxies to determine the best fit. With the accelerating generative AI race, 2025 will be a great experimental period for the intersection of AI and governance.
3. AI Agents as Proxies
Background: A more radical idea is, can we develop AI agents that understand user preferences and can vote on their behalf? Could these AI agents engage in forum discussions with other AI agents, and even co-author proposals? True AI proxies may still be far off, but next year will be a good time to explore this model.
4. Smarter Participation Incentives
Background: Experience and research show that one-time airdrop rewards are often not enough to drive long-term, deep participation in decentralized projects. In 2025, we'll see more interesting explorations "beyond airdrops". If the project needs to incentivize economically valuable activities (like transactions), direct subsidies can work; if the project wants to incentivize harder-to-measure behaviors, longer-term incentive mechanisms combined with time locks may be considered.
5. Better Funding Models for Public Goods
Background: Many projects need community members to help build the ecosystem. But existing grant and retroactive reward methods often fail to effectively incentivize high-risk, high-return ideas: members tend to choose projects with higher chances of success and getting grants, not necessarily the directions they truly believe in. A reasonable approach is to explore venture capital-like models for public goods, providing upfront funding and then giving higher rewards if the community deems the project successful. 2025 may be the "takeoff year" for such innovative funding models.
6. More Experiments with Sortition
Background: Companies like Anthropic, Meta, and OpenAI have explored using sortition (random sampling) to determine AI usage rules. Sortition involves randomly selecting a subset of users to participate in structured discussions and form collective decision recommendations. Similar experiments exist in Web3 as well. However, to truly test the viability of sortition, these citizen assemblies need to have real decision-making power, not just advisory roles, and we need to observe the effectiveness of their rulings. But for many critical issues, we need in-depth research and iterative discussions by representatives or token holders, which may not be suitable for "one-off" sortition meetings (since the assembly disbands after the discussion). Next year, we'll also see how far sortition can go.
About the Author
Andrew Hall is a professor of political economy at the Stanford Graduate School of Business and the Department of Political Science. He works at the a16z Research Lab and advises tech companies, startups, and blockchain protocols, focusing on the intersection of technology, governance, and society.
