Debate and consensus patterns in Agentic AI

Introduction

Debate and consensus patterns help multiple AI agents share ideas and agree on the best answer. This makes decisions smarter and more reliable.

When you want several AI helpers to discuss different views before choosing an answer.

When you need to reduce mistakes by having agents check each other's work.

When solving complex problems that need different opinions combined.

When building AI systems that explain their reasoning by showing a debate.

When you want AI to reach a group agreement instead of just one opinion.

Syntax

Agentic AI

debate_pattern(agents, question) -> final_answer
consensus_pattern(agents, inputs) -> agreed_answer

debate_pattern runs agents to argue different answers and picks the best.

consensus_pattern combines agent inputs to find common agreement.

Examples

Three agents debate the best movie and return the final choice.

Agentic AI

final_answer = debate_pattern([agent1, agent2, agent3], 'What is the best movie?')

Combine three agent answers to find what they all agree on.

Agentic AI

agreed_answer = consensus_pattern([agent1_output, agent2_output, agent3_output])

Sample Model

This code shows three agents giving opinions on pets. The debate pattern prints each opinion and picks the most popular. The consensus pattern finds the most common opinion from agent outputs.

Agentic AI

class Agent:
    def __init__(self, name, opinion):
        self.name = name
        self.opinion = opinion
    def argue(self, question):
        return f'{self.name} says {self.opinion}'

def debate_pattern(agents, question):
    print(f'Debate on: {question}')
    opinions = [agent.argue(question) for agent in agents]
    for op in opinions:
        print(op)
    # Simple rule: pick the opinion that appears most
    counts = {}
    for agent in agents:
        counts[agent.opinion] = counts.get(agent.opinion, 0) + 1
    final = max(counts, key=counts.get)
    return final

def consensus_pattern(agent_outputs):
    print('Consensus check on agent outputs:')
    for output in agent_outputs:
        print(output)
    # Find the most common output
    counts = {}
    for out in agent_outputs:
        counts[out] = counts.get(out, 0) + 1
    agreed = max(counts, key=counts.get)
    return agreed

# Create agents with opinions
agent1 = Agent('Alice', 'Cats are best')
agent2 = Agent('Bob', 'Dogs are best')
agent3 = Agent('Carol', 'Cats are best')

# Run debate
final_answer = debate_pattern([agent1, agent2, agent3], 'Which pet is best?')
print(f'Final answer from debate: {final_answer}')

# Run consensus on outputs
outputs = [agent1.opinion, agent2.opinion, agent3.opinion]
agreed_answer = consensus_pattern(outputs)
print(f'Agreed answer from consensus: {agreed_answer}')

Important Notes

Debate patterns help explore different views before deciding.

Consensus patterns focus on agreement, useful when agents mostly agree.

Both patterns improve AI reliability by using multiple opinions.

Summary

Debate and consensus patterns let multiple AI agents share and combine ideas.

Debate shows different opinions and picks the best one.

Consensus finds common agreement among agent answers.