Deliberative Curation: A Protocol for Multi-Agent Knowledge Bases

This paper formalizes a deliberative curation protocol for multi-agent knowledge bases. Building on a companion survey of platform governance transfer to agent settings (Johnson 2026, DOI:10.5281/zenodo.20112894), we address the gap between ephemeral multi-agent debate and persistent knowledge governance.

The protocol comprises three layers: (1) a knowledge artifact lifecycle formalized as a labeled transition system with guard conditions and timeout transitions; (2) reputation-weighted deliberative voting integrating Beta Reputation System scores with EigenTrust amplification, where structured deliberation precedes voting; and (3) graduated sanctions adapted for stateless agents, including a mechanism distinguishing technical malfunction from adversarial behavior.

We evaluate via agent-based simulation (100 agents, 7 behavioral archetypes, 2 adversity scenarios). The principal finding: the protocol trades modest precision under benign conditions for substantially better resilience under adversity (0.826 vs. 0.791 majority vote under moderate adversity, p < 0.001; 0.807 vs. 0.740 under stress). Ablation identifies commit-reveal voting as the most impactful component (+8.2–8.6pp precision). A partial open-source implementation exists.