AI-Based Instructional Strategy for Developing Communication Competencies in Requirements Elicitation Interview.

Abstract

Software requirements elicitation is a critical phase in systems development, and interviewing is the most widely used technique in professional practice. However, computer engineering students consistently face difficulties in conducting interviews, as evidenced by poorly formulated questions and a superficial exploration of stakeholders' needs. Traditional approaches have proven insufficient for developing these communication competencies. The emergence of large language models (LLMs) has opened new pedagogical opportunities by enabling the simulation of stakeholders in authentic, repeatable, and psychologically safe practice environments. This study, conducted at the University of Informatics Sciences, aimed to design and evaluate a teaching strategy based on generative artificial intelligence for the development of communication competencies in requirements elicitation interviews. A mixed-methods approach with a quasi-experimental design was employed, involving a sample of 78 third-year students distributed across four faculties. The experimental group used LLM-based simulation, while the control group employed traditional role-playing activities. Standardized instruments, including assessment rubrics, usability questionnaires, and satisfaction surveys, were applied. The results showed significant improvement in both groups, but substantially greater gains in the experimental group, with an exceptionally large Cohen's effect size. It is concluded that the integration of LLMs as simulation tools constitutes a relevant pedagogical innovation for the education of computer engineers.