Ethereum co-founder Vitalik Buterin has identified limitations of human attention as the core problem plaguing decentralized autonomous organizations and democratic governance systems.
He suggested that integrating artificial intelligence stewards could help create more efficient DAO models by enabling users to make more informed decisions.
Vitalik Buterin Proposes Using Personal LLMs to Solve Centralized Decision-Making in DAOs
In a recent X post, Buterin argued that democratic and decentralized modes of governance are marred by the limitations of human attention, because participants are faced with thousands of decisions across multiple domains of expertise without possessing sufficient time or skill to evaluate them properly.
He noted that the “usual solution” creates disempowerment as it results in a small group of delegates controlling decision-making while their supporters have zero influence once the delegate button is pressed.
According to estimates, the average participation rates in DAOs are between 15% and 25%. However, this leads to issues such as the centralization of power and ineffective decision-making. Meanwhile, worst-case scenarios can result in governance attacks, where a bad actor can acquire enough tokens to pass a malicious proposal without other members ever noticing.
Buterin proposed personal assistant large-language models (LLMs) as the solution to the human attention problem by providing users with the relevant information required for a vote. He also shared four approaches – personal governance agents, public conversation agents, suggestion markets, and privacy-preserving multi-party computation – for handling sensitive decision-making.
He said that personal governance agents can perform all the necessary actions for the user based on preferences that it infers from their personal writing, conversation history, and direct statements. If the agent is unsure about voting preferences and considers an issue important, then it would ask the user directly while providing all relevant information.
Meanwhile, public conversation agents would aggregate information from multiple DAO participants before giving each user or their LLM a chance to respond. This system would summarize individual views, convert them into shareable formats without exposing users’ private information, while identifying commonalities between inputs similar to LLM-based Polis systems – a platform designed to facilitate large-scale, AI-moderated public deliberation and consensus-building.
Ethereum Co-Founder Urges DAOs to Aggregate Collective Information Before Making Informed Decisions
Buterin highlighted that good decisions cannot come from a linear process of taking people’s views that are based only on their own information and “averaging” them. He noted that governance processes must aggregate collective information before allowing informed responses.
According to him, another challenge in a highly decentralized governance structure is when key decisions depend on private or sensitive information, such as during negotiations, internal disputes, or funding choices. Typically, DAOs solve this issue by appointing individuals who possess expertise in managing those tasks.
His proposed solution involves users being able to submit their personal LLM into a black box, where the model can view the private information, make a judgment based on that, and provide output only to that judgment. Under this framework, the user does not see the sensitive information, while no one else is able to view the contents of the user’s personal LLM, creating financial incentives for surfacing valuable contributions.
“All of these approaches involve each participant making use of much more information about themselves, and potentially submitting much larger inputs. Hence, it becomes all the more important to protect privacy,” Buterin said.
Multi-party computation using trusted execution environments could incorporate multiple individuals’ inputs without compromising privacy. Privacy protection becomes pivotal as participants submit larger inputs containing more personal information. Buterin also called for zero-knowledge proofs, which uphold anonymity, to be purpose-built into all governance tools.
At the time of writing, Ethereum (ETH) is trading at $1,884 – down 4.47% in 24 hours.
