Abstract for: A Hybrid Methodology to Bridge Expert Insight and AI Reasoning

Major public policy challenges such as climate change, population, welfare, and resource scarcity are inherently complex, shaped by nonlinear feedbacks. Thus, in developing CLD for such challenges the main question could be defined as: Which reinforcing and balancing feedbacks best explain the problem, and which policy packages could be more effective? This paper presents a new hybrid methodology to construct a Causal Loop Diagram (CLD) in a six-step process. In this methodology, Group Model Building (GMB), Large Language Models (LLMs), and Retrieval-Augmented Generation (RAG) have been integrated This methodology was applied on population dynamics and demonstrated that how combining expert insight and AI reasoning can enhance the transparency, scope, and policy relevance of systems modeling in data-limited and fast-changing contexts. This study produced two primary outcomes: (1) methodological lessons from applying LLMs in a hybrid participatory modeling process, and (2) policy insights derived from the integrated model of population dynamics. OpenAI’s ChatGPT-4o version May 2024