Shall We Team Up: Exploring Spontaneous Cooperation of Competing LLM Agents

Large Language Models (LLMs) have increasingly been utilized in social simulations, where they are often guided by carefully crafted instructions to stably exhibit human-like behaviors during simulations. Nevertheless, we doubt the necessity of shaping agents' behaviors for accurate social simulations. Instead, this paper emphasizes the importance of spontaneous Figure 1: (Depicted by GPT-4o) Two potential scenarios phenomena, wherein agents deeply engage in during a fire. People might panic and rush into crowds, contexts and make adaptive decisions without trying to exit first (left) or may stay calm, keep in line, explicit directions. We explored spontaneous and encourage others (right). In this study, we explore cooperation across three competitive scenarios whether LLM agents can simulate the gradual transition and successfully simulated the gradual emergence from non-cooperative to cooperative behaviors of agents. of cooperation, findings that align closely with human behavioral data.