Skip to content

Agent Voting

Overview

As described in What’s special about the DeGov Agent?, the DeGov Agent is an AI-powered delegate that analyzes proposals and votes on behalf of members who delegate to it. This section explains the agent voting workflow, key features, and decision process.

What is Agent Voting?

Agent Voting Process

Agent voting uses automated analysis to select an on-chain vote option. Clicking the choice button displays the agent’s commentary, with a link to detailed reasoning.

Voting Comment Preview

Decision Details

How does the agent make a vote decision?

The DeGov Agent relies on the X Interaction Model to gather and analyze both on-chain and off-chain signals:

  1. Proposal Announcement on X: When a proposal is created, the agent posts a tweet summarizing the proposal, linking to details, and launching a For/Against/Abstain poll. The poll closes before the on-chain voting deadline.
  2. Automatic Updates: The agent monitors on-chain events (status changes, votes, queueing, execution) and posts progress updates as comments on the original tweet.
  3. Community Voting: Members can vote on-chain in DeGov or off-chain via the tweet poll. Off-chain poll results do not directly affect on-chain tallies but inform the agent’s analysis.
  4. Final Decision: After the tweet poll and before on-chain voting ends, the agent synthesizes tweet poll results, comments sentiment, and live on-chain data. It then casts its on-chain vote according to the configured strategy and publishes its rationale as a comment.

Final Vote Confirmation

Voting Strategy

The DeGov Agent weights multiple inputs to ensure balanced decisions:

  1. Tweet Poll Analysis (40%): Assesses majority sentiment, turnout significance, and manipulation risks (bots, Sybil attacks).
  2. Tweet Comment Sentiment (30%): Evaluates argument quality in comments, gives extra weight to influential voices, and checks for credibility issues.
  3. On-Chain Voting Metrics (30%): Compares off-chain sentiment with on-chain vote distribution and turnout percentage to gauge engagement.
  4. Synthesis & Conflict Resolution: Combines weighted inputs for a preliminary decision. If off-chain and on-chain signals conflict sharply, the agent abstains to prompt further discussion. Otherwise, it selects the majority-backed option and assigns a confidence score (1–10).

This multi-source framework helps the agent reflect real community opinion while safeguarding against manipulation.

Last update: June 25, 2025
Created: June 25, 2025