From Text to Map: A System Dynamics Bot for Constructing Causal Loop Diagrams

We introduce and test the System Dynamics Bot, a computer program leveraginga large language model to automate the creation of causal loop diagrams fromtextual data. To evaluate its performance, we ensembled two distinct databases.The first dataset includes 20 causal loop diagrams and associated texts sourcedfrom the system dynamics literature. The second dataset comprises responsesfrom 30 participants to a vignette, along with causal loop diagrams coded bythree system dynamics modelers. The bot uses textual data and successfullyidentifies approximately sixty percent of the links between variables andfeedback loops in both datasets. This paper outlines our approach, providesexamples, and presents evaluation results. We discuss encountered challengesand implemented solutions in developing the System Dynamics Bot. The bot canfacilitate extracting mental models from textual data and improve modelbuilding processes. Moreover, the two datasets can serve as a testbed forsimilar programs.