Elicitação aprimorada de requisitos de software por meio da integração do método 5W2H com IA generativa

Abstract

Requirements Engineering plays a fundamental role in software development, as deficiencies in requirements elicitation and specification can lead to rework, increased costs, and a higher risk of project failure. Structured methods, such as 5W2H, support problemunderstanding but still rely heavily on human expertise. In this context, Generative Artificial Intelligence (Generative AI) has emerged as a promising alternative, contributing to the reduction of ambiguities and the partial automation of requirements elicitation and specification activities. This study proposes the development and evaluation of a methodology that integrates the 5W2H method with Generative AI to improve the quality of both functional and non-functional requirements, considering quality criteria established in the literature. In addition, it compares the proposed integrated approach with the traditional requirements elicitation process. The research was structured according to the experimental process described by Wohlin et al. (2012), comprising the following phases: (i) scoping, (ii) planning, (iii) operation, (iv) analysis and interpretation of results, and (v) presentation and packaging. The proposed approach enabled the production of higher-quality requirements that satisfy the quality criteria recommended in the literature, while also reducing the time required for requirements development and providing greater cognitive support to participants. Furthermore, the study discusses the limitations and risks associated with the use of Generative AI, such as bias and overreliance, reinforcing the importance of human oversight and contributing an empirically evaluated methodologyfor software requirements elicitation and specification.