Effective Prompting with ChatGPT in Engineering Optimization

This article is a research paper that discusses the application of ChatGPT in the formulation of optimization problems within engineering optimization. It highlights the challenges faced by practitioners in creating effective optimization problem formulations, an intricate process that is often time-consuming and complex. The authors evaluate the efficacy of various self-designed prompts used with different ChatGPT models, focusing on the significance of specificity in word choice and the use of domain-specific terminology. The research employs statistical methods, including analysis of variance and Tukey’s test, to assess the impact of prompt wording on the quality of optimization solutions. Additionally, the paper investigates the sequential learning approach to enhance responses from ChatGPT. The findings indicate that carefully selected wording can improve problem formulations and that sequential learning significantly contributes to the overall quality of optimization problem definitions. This work aims to increase interest in utilizing ChatGPT for problem formulation in engineering contexts.