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

We had an insightful webinar titled “From Text to Map: A System Dynamics Bot for Constructing Causal Loop Diagrams.” The session, led by Niyousha Hosseinichimeh, explored the capabilities and challenges of a novel tool designed to automate the creation of causal loop diagrams (CLDs) from textual data. Here are the key takeaways from the presentation:

Overview of the System Dynamics Bot

The System Dynamics Bot aims to automate the process of constructing CLDs from text, leveraging large language models and generative artificial intelligence. Niyousha Hosseinichimeh emphasized the tool’s potential by stating, “The System Dynamics Bot can streamline the modeling process, making it less labor-intensive and more efficient.”

1. Capabilities of Large Language Models: The bot uses deep learning techniques to process and interpret text, identifying causal relationships and constructing CLDs. This demonstrates the potential of large language models in system dynamics and model building.
2. Effectiveness and Practical Examples: Practical examples during the webinar showcased the bot’s ability to convert textual data into causal loop diagrams, effectively capturing significant feedback loops and causal links. Hosseinichimeh noted, “This tool offers a novel way to enhance our understanding of complex systems by automating the creation of causal loop diagrams.”

Evaluation of the Bot’s Performance

Two data sets were used for evaluation:

1. Data Set One: Included 20 CLDs from system dynamics literature, capturing 60% of causal links and 66% of feedback loops on average.
2. Data Set Two: Included responses from 30 individuals, with the bot capturing 56% of causal relationships and 83% of feedback loops.

Applications and Future Directions

The bot can significantly save time and effort in developing CLDs, assisting both novice modelers and experienced practitioners. Potential applications include:

• Educational Tools: Enhancing the learning experience for system dynamics students.
• Group Model Building: Supporting the creation of causal loop diagrams during group sessions.
• Comparative Studies: Comparing human-generated and AI-generated models to improve accuracy and insights.
• Literature Review-Based CLDs: Synthesizing information from various fields to create comprehensive diagrams.
• Mapping Mental Models: Analyzing how different individuals perceive complex issues.

Hosseinichimeh highlighted, “This program can be used to generate literature review-based causal loop diagrams and aid in group model building sessions,” showcasing its versatility and practical benefits.

Watch the full webinar recording below. It provides in-depth insights into the development, functionality, and applications of the System Dynamics Bot, showcasing its potential to revolutionize how we construct and interpret causal loop diagrams.

Watch the recording below

PRESENTERS

Niyousha Hosseinichimeh has a PhD and a master’s degree in public administration and policy from State University of New York at Albany, and a bachelor’s degree in mechanical engineering from Sharif University, Iran. She is currently an assistant professor at the Department of Industrial and Systems Engineering at Virginia Tech. Her research focuses on developing and applying methods to improve health and healthcare systems. She uses simulation models to help stakeholders improve their understanding and decision making in complex dynamic systems. She has applied system dynamics approach to diverse health issues including infant mortality, mental health, and alcohol impaired driving among teens. Her methodological contributions include expanding calibration methods for dynamic models and developing techniques for system dynamics group model building. Her research has been funded by the National Institute of Health, National Science Foundation, Agency for Healthcare Research and Quality, Ohio State Department of Health, and Burroughs Wellcome Fund.