Abstract for: CHIRON: Integrating AI and Rigorously Interpreted Quotation Analysis to Enhance Conceptualization in System Dynamics

Significant work has been accomplished on the formalization and automation of text analysis methods for system dynamics in the conceptualization phase. Large Language Models (LLMs) offer the potential to empower modelers to further reduce the effort required to conduct this empirically rigorous approach. We use a framework to describe the theoretical and practical human and LLM roles in a hybrid text analysis and LLM approach for SD simulation studies. Causal Hybrid artificial Intelligence and Rigorously interpreted quotatiON dynamic analysis process (CHIRON), is a framework that integrates AI and SD using a formal approach to text analysis in model development, evaluation and documentation. Using proposed roles for human and LLM in a simplified project model framework, we describe our progress to date in advancing an AI-LLM integration in our SD research. Garry Kasparov likened the human-AI hybrid to a centaur. In mythology, centaurs are wild, except for Chiron, a wise mentor. Similarly, the CHIRON framework highlights an LLM’s mentor-like role when trained on rigorously curated SD text documentation. This approach outperforms human-only methods by handling time-consuming tasks, managing its limitations effectively, and demonstrating built-in wisdom through its humble recognition of limits. Our team's expertise in text analysis for SD modeling has guided the design of the CHIRON framework, revealing AI's strengths and limitations in SD. Rigorous documentation shapes the training data needed for an effective hybrid AI-SD process. Our experience highlights the value of this approach, maximizing the LLM’s potential by providing greater bandwidth and preventing it from being overwhelmed by complex data. GPT 4.0