This dissertation is accessible only to the Illinois State University community.

  • Off-Campus ISU Users: To download this item, click the "Off-Campus Download" button below. You will be prompted to log in with your ISU ULID and password.
  • Non-ISU Users: Contact your library to request this item through interlibrary loan.

Graduation Term

Spring 2026

Degree Name

Master of Science (MS)

Department

Department of Technology

Committee Chair

Borinara Park

Committee Member

Sundeep Inti

Abstract

Causal loop diagrams (CLDs) capture cause relationships and feedback structures but remain qualitative and cannot support dynamic simulation. For quantitative or policy analysis, CLDs must be converted into stock and flow diagrams (SFDs). Several methods, such as labeling, annotated CLD (aCLD), and group model building (GMB) still rely heavily on expert judgment and repeated discussions. This study applies the AI Agent Modular Capabilities (AAMC) framework to assess the feasibility of automating CLD construction and CLD-to-SFD conversion and to identify methods capable of producing structurally sound SFDs. Using a tariff-related case study, a CLD was generated through triangulation and then converted into multiple SFDs using labeling, aCLD, and GMB. Initial results show that although all models preserved the main feedback loops, each contained structural issues. After refining the conversion instructions, a second analysis found that the aCLD model provided greater informational completeness. Comparing model combinations further shows that the labeling + aCLD approach maintained logical consistency. A validation confirms that both aCLD and labeling + aCLD perform consistently in the AI energy demand scenario. Overall, aCLD is effective in producing models that contain more complete structural information, while the combined approach offers clearer and interpretable structures.

Access Type

Thesis-ISU Access Only

Download

Available for download on Thursday, July 15, 2027

Off-Campus Download

Share

COinS