Collaborative Spatial Learning with Multi-LLM Agents in Networked Social Experiments

2026-07-16

AI Summary

This study investigates how communication network structure influences the collective problem-solving efficiency of multi-agent systems in spatial search tasks, with a focus on the exploration–exploitation trade-off. The authors deploy sixteen large language model (LLM) agents across eight Mason–Watts networks to replicate human social experiments, using mechanistic Bayesian optimization agents as a benchmark. They systematically evaluate the effects of network path length and initialization strategies on collaborative performance. This work presents the first application of LLMs in networked collective intelligence experiments, revealing that simple instructions—such as randomization in the initial round—can substantially amplify network efficiency effects, yielding collective gains more than three times greater than those attributable to topological differences alone. Furthermore, the study establishes a multidimensional comparison framework among LLMs, Bayesian agents, and human behavior, uncovering quantifiable differences in exploration strategies, imitation tendencies, and spatial diversity.

Abstract

Collective problem solving often requires that group members consider the tradeoff between exploitation of known solutions and exploration for new ones, where information of known solutions can be disseminated among individual members through communication networks. The Mason--Watts experiment (PNAS 2012) showed that human groups in shorter-path networks outperform those in longer-path networks on a two-dimensional search task. In this work, we focus on the investigation of such network-efficiency effects in the setting of a group of large language model (LLM) agents. Specifically, we consider groups of sixteen LLM agents playing the Mason--Watts experiment on the eight Mason--Watts network topologies. Moreover, we develop mechanistic Bayesian optimization agents such that the performance of LLM agents can be compared with both the mechanistic agents and the human experimental data. Our computational experiments indicate that the LLM agents show a significant network-efficiency effect when instructed to randomize their first-round choices, but not under the default initialization. In this experiment, adding a one-sentence first-round randomization instruction improves collective payoff by more than three times the estimated payoff difference across the eight network topologies. Also, the Bayesian optimization agents obtain higher payoffs than the evaluated LLM agents on this spatial search task. We further compare the agents' exploration--exploitation behavior, copying, and spatial diversity.

Problem

• network-efficiency effect
• collective problem solving
• spatial search task
• exploration-exploitation tradeoff
• multi-LLM agents

Innovation

• multi-LLM agents
• networked social experiments
• spatial search
• Bayesian optimization agents
• exploration-exploitation tradeoff