Abstract for: Inception of Thought: AI Extracting Mental Models from System Dynamics Society Presidents

This research addresses how to systematically extract mental models of System Dynamics Society (SDS) presidential speeches to uncover evolving perspectives on education and quality. By leveraging Generative AI, specifically ChatGPT-4, this study introduces a structured prompting method to extract key themes, feedback loops, and drawing CLDs. This AI-assisted approach offers insights into how leadership perspectives can impact trajectory of the System Dynamics Society. This study applies ChatGPT-4o using a structured prompting method based on the CREATE Framework to extract themes, identify feedback loops, and construct Causal Loop Diagrams (CLDs). AI-generated insights are validated manually, with CLDs built using Stella Architect. Speeches from Forrester, Zahn, Milling, and Saeed are used as they are foundational references. Outputs are structured into Excel tables, ensuring accuracy and consistency analyzing SDS presidential discourse. The AI-assisted analysis of SDS presidential speeches from 1983 to 1995 revealed an evolving focus on education and quality, transitioning from methodological rigor (Forrester) to policy integration (Saeed). AI-extracted feedback loops and Causal Loop Diagrams (CLDs) highlighted systemic relationships. However, AI faced challenges in recognizing nested and emergent loops, requiring manual validation to refine assumptions and ensure the accuracy and coherence of leadership discourse. AI-assisted analysis enhances System Dynamics research by extracting themes and feedback loops, informing direction of education and quality. However, human validation is essential to refine nested loops and systemic interactions. Future research should improve AI’s ability to recognize complex feedback loops, expand into other presidential discourses, and refining prompting methods to obtain better results. This study highlights AI’s role in supporting, not replacing, expert-driven analysis. AI extracted themes, structured results, and assisted writing with human validation.