Primary aims

Contents

Primary aims#

Our study investigates the feasibility of using generative AI to recreate DES models in healthcare based on textual descriptions from the academic literature. We focus on generating models in the Python simulation package SimPy [Team SimPy, 2020], selected for its (i) compatibility with language models’ code-generating capabilities, (ii) growing adoption in health service Operational Research [], and (iii) our expertise in developing SimPy models for healthcare applications [].

To assess feasibility, we engineer prompts for Perplexity.AI to generate complete Python and SimPy code that captures model logic (e.g. arrival processes, queuing, activities, sampling, and balking). Additionally, we explore generating browser-based user interfaces using Streamlit [] to enhance accessibility for non-programmers. Our research objectives are to:

  • Determine if generative AI can produce functional, verifiable SimPy models from engineered prompts describing DES models

  • Assess the feasibility of generating usable Streamlit web interfaces for these models

  • Pilot this approach by recreating two published healthcare DES models

  • Evaluate the reproducibility of our methodology when conducted by different modelers

This work contributes to the growing interest in generative AI applications for modeling [Frydenlund et al., 2024, Giabbanelli et al., 2024, Giabbanelli, 2024, Tolk, 2024]. Our long-term goal is to develop guidance on prompt engineering and to document the opportunities, challenges, and limitations of using AI to recreate DES models—ultimately supporting result reproduction, model reuse, and educational applications.

References#

[1]

Erika Frydenlund, Joseph Martínez, Jose J Padilla, Katherine Palacio, and David Shuttleworth. Modeler in a box: how can large language models aid in the simulation modeling process? Simulation, pages 00375497241239360, 2024.

[2]

Philippe J Giabbanelli, Jose J Padilla, and Ameeta Agrawal. Broadening access to simulations for end-users via large language models: challenges and opportunities. In 2024 Winter Simulation Conference (WSC), 2535–2546. IEEE, 2024.

[3]

Philippe J. Giabbanelli. Gpt-based models meet simulation: how to efficiently use large-scale pre-trained language models across simulation tasks. In Proceedings of the Winter Simulation Conference, WSC '23, 2920–2931. IEEE Press, 2024.

[4]

Andreas Tolk. Hybrid modeling integrating artificial intelligence and modeling & simulation paradigms. In 2024 Winter Simulation Conference (WSC), 1271–1280. IEEE, 2024.

[5]

Team SimPy. Simpy 3.0.11. https://simpy.readthedocs.io/en/latest/index.html, 2020. Accessed: 2023-10-18.

Contents