Consensus in Multi-Agent Systems

Master decision-making strategies when agents disagree

Weighted Voting & Expert Influence

Not all agent opinions are equally valuable. An expert agent with deep domain knowledge should have more influence than a novice agent. Weighted voting assigns different vote strengths based on expertise, experience, or reliability.

When to Use Weighted Consensus

Domain Expertise Matters: Medical diagnosis, legal analysis, technical architecture decisions
Track Record Exists: Agents have measurable accuracy/success rates from past decisions
Responsibility Differs: Some agents bear more consequences from wrong decisions

Interactive: Weight Adjuster

Adjust agent weights and preferences to see how weighted voting can produce different results than simple majority voting.

SET AGENT WEIGHTS & PREFERENCES

🎓
Expert Agent
Weight: 3
15
💼
Experienced Agent
Weight: 2
15
👶
Junior Agent
Weight: 1
15

UNWEIGHTED (1 agent = 1 vote)

OPTIONA2 votes
OPTIONB1 votes
OPTIONC0 votes
Winner
OPTIONA

WEIGHTED (by expertise)

OPTIONA4 points
OPTIONB2 points
OPTIONC0 points
Winner (67% of total)
OPTIONA

Determining Agent Weights

Common Weighting Strategies

1.
Historical Accuracy: Weight based on past prediction success rate
2.
Domain Expertise: Higher weights for specialized knowledge areas
3.
Confidence Scores: Agents self-report certainty levels
4.
Stake/Risk: Those bearing consequences get more say

Example: Medical Diagnosis

🩺 Specialist DoctorWeight: 5
👨‍⚕️ General PractitionerWeight: 3
🤖 AI Diagnostic ToolWeight: 2
📚 Medical StudentWeight: 1

💡 Key Insight

Weighted voting respects expertise without creating dictatorships. The expert agent doesn't unilaterally decide, but their opinion carries more weight. This balances democratic participation with recognition that some agents truly know more. The challenge is setting fair, justifiable weights.