Lowering the Barrier: Embedding the Modeling Process Into AI Agents
System Dynamics Review · 2026-10 · 42 (4)
DOI: 10.1002/sdr.70046 ↗Kurz gefasst
Fachartikel in der Zeitschrift System Dynamics Review, online erschienen am 17.11.2026. Verfasst von William Schoenberg. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
This paper presents an agentic artificial intelligence (AI) architecture that embeds the System Dynamics modeling process within a large language model driven workflow. Building on earlier work that used language models to generate and interpret simulation models, the architecture extends AI support from single‐step interactions to iterative, tool‐mediated modeling activities. The system integrates model construction, simulation, validation, feedback analysis, calibration, sensitivity analysis, and visualization within a unified agent loop. Rather than requiring users to manually coordinate these tasks, the agent conducts iterative cycles of model development, testing, and refinement while maintaining explicit, simulatable model artifacts. A worked example demonstrates how the architecture supports the complete modeling lifecycle, from problem formulation through calibration and uncertainty analysis. By grounding outputs in transparent simulation models and exposing intermediate analytical results, the approach aims to reduce procedural barriers to System Dynamics practice while preserving opportunities for inspection, learning, validation, and methodological rigor.
Zitieren
APA
Schoenberg, W. (2026). Lowering the Barrier: Embedding the Modeling Process Into AI Agents. System Dynamics Review, 42(4). https://doi.org/10.1002/sdr.70046
Vancouver
Schoenberg W. Lowering the Barrier: Embedding the Modeling Process Into AI Agents. System Dynamics Review. 2026;42(4). doi:10.1002/sdr.70046
BibTeX
@article{Schoenberg2026,
author = {Schoenberg, William},
title = {Lowering the Barrier: Embedding the Modeling Process Into AI Agents},
journal = {System Dynamics Review},
year = {2026},
volume = {42},
number = {4},
doi = {10.1002/sdr.70046}
}
EUVIMED