Abstract for: Systematic Variable Classification in Causal Loop Diagram Formalization: A Multi-Method Labeling Procedure
Causal loop diagrams (CLDs) are widely used to map complex health problems. While translating CLDs into system dynamics models (SDMs) can provide valuable insights, it requires several steps that not often taken—including the classification of variables into stocks, flows, auxiliaries, and constants, a process known as labeling. Despite its importance, there is no universally accepted method for systematically distinguishing between these categories. Existing approaches tend to rely on modeling intuition and implicit assumptions, leading to inconsistencies, limited reproducibility, and barriers for non-modelers. In this paper, we propose a structured procedure that integrates multiple labeling methods, documents uncertainties, and seeks to enhances stakeholder involvement. Our approach incorporates the input of experts and large language models through a guided questionnaire, leveraging analogy-based reasoning, structural heuristics, and response timescales. We apply the approach to a use-case causal loop diagram on sleep and mental health problems in young adults. By making the labeling process more robust and transparent, this approach aims to improve CLD-to-SDM formalization, ultimately supporting more rigorous and accessible modeling practices in health research and beyond. ChatGPT was used to answer the questionnaire