Generative AI for Sustainable Activity Diagram Modelling: The Role of Case Complexity and Prompt Language

Abstract

Sustainable software engineering prioritizes efficiency, minimizes resource use, and streamlines development methodologies. In software modelling, sustainable methodologies seek to reduce cognitive load and extraneous effort while enhancing learning outcomes. This study examines the function of Generative Artificial Intelligence (AI) as a sustainable technology in facilitating UML activity diagram modelling. A quasi-experimental study utilizing a repeated-measures methodology was performed with 30 undergraduate students. The diagrams were evaluated via an analytical rubric. The data were examined utilizing descriptive statistics, paired sample t-tests, and repeated-measures ANOVA. The findings indicated a substantial enhancement in diagram quality when participant utilised Generative AI in contrast to the non-AI condition (t(29) = -14.31, p < 0.001, d = 2.61). A notable disparity was seen between simple and complex situations (t(29) = 2.94, p = 0.006, d = 0.54). No substantial difference was observed between Indonesian and English prompts. A notable interaction impact was identified between Generative AI and case complexity (F(1,29) = 17.96, p < 0.001, η 2 = 0.38). From a sustainability standpoint, these findings suggest that Generative AI improves modelling efficiency, minimises redundant cognitive labour, and facilitates more resource-efficient learning methodologies. This study emphasises the capability of Generative AI as a resource for sustainable software modelling in educational and engineering contexts.