Abstract for: How do we learn from system dynamics models

This paper addresses how learning from system dynamics (SD) models is conceptualized, operationalized, and measured across diverse research traditions. Despite widespread use of models as learning tools, the literature remains fragmented, with inconsistent definitions and evaluation approaches. This review seeks to clarify what it means to learn from models and why this matters for improving decision-making in complex systems. The study conducts a structured literature review of 54 empirical and conceptual papers published between 2010 and 2026. It organizes the field along three dimensions: conceptualizations of learning (cognitive, behavioral, social), engagement modes (model construction, simulation interaction), and measurement approaches. An iterative coding framework enables systematic comparison across studies and identification of patterns and gaps. The review finds that learning is conceptualized in three primary ways: changes in mental models, improvements in decision performance, and development of shared understanding. However, these perspectives are rarely integrated. Results also show that engagement mode shapes learning processes, and that measurement approaches are highly varied, limiting comparability and making it difficult to draw cumulative conclusions. The findings highlight the need for stronger alignment between learning theories, modeling approaches, and evaluation methods. For researchers and practitioners, this implies designing interventions more deliberately and selecting measures that match intended outcomes. Advancing this alignment can improve both the rigor of SD research and the practical effectiveness of models in supporting real-world decision-making. Editing and initial search for papers to include