NetworkGames: Simulating Cooperation in Network Games with Personality-driven LLM Agents
Xuan Qiu
TL;DR
This work introduces NetworkGames, an open-source framework that uses MBTI-based personality prompts to drive LLM agents embedded in network topologies, enabling the study of cooperation dynamics in Iterated Prisoner's Dilemma settings. It conducts exhaustive dyadic experiments across 16 personality types to build micro-level interaction matrices and then examines macro-level emergence in three network topologies, revealing that topology and personality distribution jointly determine cooperation levels. Key findings include a robust pro-social MBTI cluster that cooperates heavily, a paradoxical fragility of Small-World networks to defection cascades, and substantial gains in cooperation when pro-social hub nodes are strategically placed in Scale-Free networks. The results have practical implications for designing healthier online environments and demonstrate the value of an open-source, personality-driven simulation framework for network game research.
Abstract
The advent of Large Language Models (LLMs) presents a novel opportunity to build high-fidelity agent-based models for simulating complex social systems. However, the behavior of these LLM-based agents in game-theoretic network games remains surprisingly unexplored. In this work, we introduce "NetworkGames," a novel simulation framework designed to investigate how network topology and agent personality jointly shape the evolution of cooperation in network games. We instantiate a population of LLM agents, each endowed with a distinct personality from the MBTI taxonomy, and situate them in various network structures (e.g., small-world and scale-free). Through extensive simulations of the Iterated Prisoner's Dilemma, we first establish a baseline dyadic interaction matrix, revealing nuanced cooperative preferences between all 16 personality pairs. We then demonstrate that macro-level cooperative outcomes are not predictable from dyadic interactions alone; they are co-determined by the network's connectivity and the spatial distribution of personalities. For instance, we find that small-world networks are detrimental to cooperation, while strategically placing pro-social personalities in hub positions within scale-free networks can significantly promote cooperative behavior. Our findings offer significant implications for designing healthier online social environments and forecasting collective behavior. We open-source our framework to foster further research in network game simulations.