Dynamic Consensus System — multi-agent belief, lock accumulation, consensus.
When agents disagree, how do they reach consensus? Plato DCS implements a dynamic consensus system where agents accumulate locks on beliefs until a threshold is reached.
- Propose — An agent proposes a belief
- Accumulate — Other agents lock their agreement
- Threshold — When enough agents agree, the belief is accepted
- Dynamic — Thresholds adjust based on confidence and agent expertise
- Multi-agent belief tracking — Each agent's beliefs are recorded
- Lock accumulation — Agreement is collected, not voted
- Dynamic thresholds — Consensus requirements adjust based on stakes
- Conflict resolution — When agents disagree, evidence wins
pip install plato-dcsEnables democratic decision-making across the fleet without a central authority.
MIT