Collective agent behavior studies the emergent phenomena that arise when networks of LLM-powered agents interact, communicate, and make decisions together. Unlike classical agent-based models using simple rules, cognitive agents powered by LLMs exhibit advanced reasoning that fundamentally changes how consensus, polarization, and swarm dynamics emerge at scale.

Classical collective behavior models (particle swarm optimization, opinion dynamics, Schelling segregation) use agents governed by formal mathematical rules. LLM agents differ in three critical ways:

• Reasoning: They interpret context, weigh arguments, and generate novel strategies
• Communication: They exchange natural language, not just numerical signals
• Bias: They carry training data biases that create systematic behavioral patterns

The central question is whether embedded “intelligence” improves or degrades collective outcomes compared to simple rule-following particles.

De Marzo, Castellano, and Garcia published “Unraveling the emergence of collective behavior in networks of cognitive agents” in npj Artificial Intelligence (March 2026), comparing LLM agent swarms with classical particles across two tasks.

LLM Agent Swarm Optimization (llmASO): A swarm of interacting LLM agents acts as a function optimizer. Each agent proposes solutions, observes neighbors' results, and updates its strategy through natural language reasoning.

Key findings:

• Individual LLM agents outperform particles in decision-making quality
• However, consensus tendencies make swarms prone to premature convergence — agents agree too quickly on suboptimal solutions
• Adjusting network topology (reducing connectivity) alleviates premature convergence but slows overall convergence
• In the Schelling segregation model, local interactions and homophilic mechanisms produce self-organized spatial patterns similar to human segregation dynamics

$$F_{majority} = \frac{n_{agree}}{n_{total}} \cdot w_{conformity}$$

The majority force coefficient governs how strongly agents conform to neighborhood consensus. Coordination breaks down beyond a critical group size $N_c$ that grows exponentially with model language ability.

LLM agents exhibit conformity bias consistent with Social Impact Theory from social psychology. Conformity pressure depends on:

• Group size: Larger unanimous groups exert stronger influence
• Unanimity: A single dissenter dramatically reduces conformity
• Source authority: Agents defer more to responses framed as authoritative

Even high-performing agents in isolation falter under peer pressure near their competence boundaries. This has implications for multi-agent systems where agents vote or debate — the “wisdom of crowds” effect can reverse into groupthink.

# Simulating opinion dynamics in an LLM agent network
import networkx as nx
import numpy as np
 
class CognitiveAgent:
    def __init__(self, agent_id, initial_opinion, stubbornness=0.3):
        self.id = agent_id
        self.opinion = initial_opinion  # continuous value in [0, 1]
        self.stubbornness = stubbornness
 
    def update_opinion(self, neighbor_opinions):
        compatible = [o for o in neighbor_opinions
                      if abs(o - self.opinion) < 0.3]  # confidence threshold
        if compatible:
            mean_neighbor = np.mean(compatible)
            self.opinion = (
                self.stubbornness * self.opinion +
                (1 - self.stubbornness) * mean_neighbor
            )
        return self.opinion
 
def simulate_opinion_dynamics(n_agents=50, n_steps=100, topology="watts_strogatz"):
    if topology == "watts_strogatz":
        G = nx.watts_strogatz_graph(n_agents, k=6, p=0.3)
    elif topology == "barabasi_albert":
        G = nx.barabasi_albert_graph(n_agents, m=3)
    else:
        G = nx.complete_graph(n_agents)
 
    agents = {i: CognitiveAgent(i, np.random.random()) for i in range(n_agents)}
    history = []
    for step in range(n_steps):
        for node in G.nodes():
            neighbors = list(G.neighbors(node))
            neighbor_opinions = [agents[n].opinion for n in neighbors]
            agents[node].update_opinion(neighbor_opinions)
        history.append([agents[i].opinion for i in range(n_agents)])
    return np.array(history)

In resource-constrained environments, diverse LLM agent populations (GPT-2, OPT, Pythia) spontaneously polarize into tribes:

• Followers (high-p): Agents that conform to majority behavior, achieving 84% individual success
• Anti-followers (low-p): Contrarian agents that deliberately diverge

Under scarcity ($C/N = 0.14$), tribal formation reduces demand variance and enables individual wins despite 92% system overload. Under abundance ($C/N = 0.86$), the same tribal structures hinder equitable resource distribution.

The optimal topology depends on the task. For optimization, sparse networks preserve diversity. For coordination tasks requiring agreement, denser networks accelerate consensus.

• Guardrails against groupthink: Introduce structured dissent or temperature variation across agents
• Topology-aware orchestration: Design communication graphs to match task requirements
• Diversity by design: Mix model families and prompting strategies to prevent correlated failures
• Scale awareness: Monitor for phase transitions where collective behavior qualitatively changes

• Unraveling collective behavior in networks of cognitive agents (de Marzo et al., npj AI, 2026)
• LLM Agents as Strategies in Social Dilemmas at Scale (2026)
• Motbook: Social Dynamics of AI Agent Populations (2026)
• Symmetry Breaking in Collective Decision-Making (npj Complexity, 2026)
• GenSwarm: Multi-Robot Code-Policy Generation via LLMs (npj Robotics, 2026)

• Agent Personalization
• Multimodal Agent Architectures
• Agent Cost Optimization