Decentralized Metaverse Governance via AI-Supported Voting Models

Abstract

The advent of the metaverse—a collection of persistent, shared, 3D virtual spaces—promises unprecedented opportunities for social interaction, commerce, and digital creativity. However, as virtual communities scale, the need for effective governance mechanisms becomes paramount. Traditional centralized models suffer from single points of failure, censorship risk, and lack of transparency, while purely token-weighted decentralized autonomous organizations (DAOs) often see low participation, plutocratic voting dynamics, and vulnerability to sybil attacks. To address these challenges, we introduce Decentralized Metaverse Governance via AI-Supported Voting Models (DMG-AI), a hybrid framework that marries blockchain-based voting with machine-learning–driven anomaly detection. In DMG-AI, voting rights are allocated via on-chain tokens, but each vote is evaluated off-chain by an AI module trained to identify patterns indicative of manipulative behaviors—such as vote-buying, collusion, or sybil identity proliferation. We built a prototype on a private Ethereum testnet and conducted a mixed-methods evaluation comprising an online user survey (N=250) and agent-based simulations (1,000 agents, 100 proposals). Survey results show an 18 percentage-point increase in voter turnout and a 0.9-point rise in perceived governance trust (on a 1–5 Likert scale) under DMG-AI versus baseline DAOs. Simulation experiments demonstrate a 67% reduction in successful sybil attacks and achieve 85% precision and 78% recall in anomaly detection, with a modest 1-hour increase in proposal resolution time. These findings confirm that integrating AI into decentralized governance enhances both security and inclusivity for metaverse communities.