Generative Agent-Based Modeling: Unveiling Social System Dynamics through Coupling Mechanistic Models with Generative Artificial Intelligence

Navid Ghaffarzadegan,Aritra Majumdar,Ross Williams,Niyousha Hosseinichimeh

DOI: https://doi.org/10.1002/sdr.1761

2023-09-21

Abstract:We discuss the emerging new opportunity for building feedback-rich computational models of social systems using generative artificial intelligence. Referred to as Generative Agent-Based Models (GABMs), such individual-level models utilize large language models such as ChatGPT to represent human decision-making in social settings. We provide a GABM case in which human behavior can be incorporated in simulation models by coupling a mechanistic model of human interactions with a pre-trained large language model. This is achieved by introducing a simple GABM of social norm diffusion in an organization. For educational purposes, the model is intentionally kept simple. We examine a wide range of scenarios and the sensitivity of the results to several changes in the prompt. We hope the article and the model serve as a guide for building useful diffusion models that include realistic human reasoning and decision-making.

Artificial Intelligence,Machine Learning,Multiagent Systems,Adaptation and Self-Organizing Systems,Physics and Society

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use Generative Artificial Intelligence (GAI) and Large Language Models (LLMs) to more effectively represent individual decision - making behaviors in social systems, especially by combining these models with Agent - Based Modeling (ABM). Traditional social system modeling methods face challenges in representing human decision - making because these models usually assume that humans are completely rational profit maximizers or rely on simplified decision - making rules. However, in reality, the decision - making process is often influenced by factors such as limited information, short - sighted behavior, different decision - making heuristics, and satisfying rather than maximizing preferences. Therefore, these traditional models may not be able to accurately capture real - life human behavior patterns.

To solve this problem, the paper proposes the concept of Generative Agent - Based Modeling (GABM). The core idea of GABM is to build models at the individual level, with each agent (i.e., the individual in the model) connected to a large - language model, enabling LLMs - driven decision - making. This method allows agents to reason and make decisions based on the state of the system and the behavior of other agents without the model builder pre - setting specific decision - making rules. In this way, GABM aims to provide a new method for developing dynamic models of social systems that incorporate more realistic human reasoning and decision - making.

Specifically, the paper demonstrates how to implement GABM through a simple case study - a diffusion model of dress code in an office. In this model, 20 employees choose to wear either a green or a blue shirt to work every day. The model uses ChatGPT (GPT - 3.5 - Turbo version) as the LLM, and each agent will send a request to the LLM at each step based on the environmental information and personal characteristics of the previous day to decide which color shirt to wear that day. In this way, the model can not only simulate the diffusion process of norms but also explore the influence of different personality characteristics on individual decision - making.

In conclusion, the main objective of this paper is to explore how to improve social system modeling by combining LLMs and ABM, especially to increase the authenticity and complexity of human behavior representation in the model.