Problem

The active involvement of users and customers in agile software development often presents practical challenges. Furthermore, with the rise of Generative AI (GenAI), there's a need to integrate this technology meaningfully into higher education, ensuring students understand its applications, limitations, and ethical implications. This paper addresses these issues by proposing a method to teach Agile Requirements Engineering (RE) that familiarizes students with GenAI tools in a pedagogically sound manner.

Method

The authors developed a stakeholder simulation for Agile RE education. Students are provided with a meta-prompt, which they use with a GenAI chatbot (e.g., ChatGPT) to simulate AI Personas representing different stakeholders. Through interviews with these AI Personas, students elicit requirements. Following the interviews, students apply agile practices like story mapping or impact mapping to document the gathered requirements. The process concludes with a structured group discussion where students critically reflect on their experiences, the quality of artifacts, and the technical and ethical limitations of using GenAI, including biases in training data and data privacy. The meta-prompt approach ensures flexibility and independence from specific LLM providers.

Results

This teaching approach provides students with practical experience in state-of-the-art agile practices for requirements elicitation and documentation. Concurrently, they develop a crucial understanding of the technical and ethical limitations associated with generative AI. The paper notes experiences with different agile practices: for user stories, students initially generated superficial outputs needing lecturer guidance for refinement; for story maps, novice prompting led to feature collections rather than user-centered narratives; and for impact maps, students struggled with granularity and distinguishing stakeholders from actors. The simulation initially used customized-GPT but transitioned to a meta-prompt due to licensing and data protection concerns, allowing students to choose their GenAI tool. Optimization of the meta-prompt led to more efficiency-oriented, less human-like AI Persona interactions, an area for future refinement.

Implications

The integration of AI Personas into Agile RE education is shown to be a valuable pedagogical instrument when applied in a structured and reflective manner. This approach prepares future requirements engineers for real-world challenges while equipping them with critical thinking skills regarding emerging technologies. Future work will focus on refining meta-prompts to enhance the human-likeness and interactivity of AI Personas, evaluating agentic AI, and exploring alternative interaction forms like voice interfaces. The authors also plan systematic evaluation of learning outcomes and extensions of the teaching case to other software engineering and project-based learning contexts.