Abstract for: Simulation and ChatGPT: Large Language Models as Tools for Model Analysis

This paper explores the potential of large language models (LLMs), particularly ChatGPT, in supporting system dynamics (SD) practitioners in model analysis. Amidst growing public interest and concerns surrounding LLMs, we propose leveraging ChatGPT's capabilities to interpret existing causal loop diagrams and forecast system behavior. Through a comparative analysis between traditional simulation techniques and ChatGPT forecasts, we examine the benefits and limitations of employing LLMs in SD. While ChatGPT demonstrates promise in understanding model structures and facilitating communication with stakeholders, it also reveals shortcomings in grasping nuanced system behaviors and may produce misleading outputs. Our findings underscore the importance of utilizing LLMs as supplementary and educational tools rather than replacements for established simulation methods.