Teaching Agile Requirements Engineering: A Stakeholder Simulation with Generative AI
A new teaching method uses GenAI personas to simulate real-world stakeholder interviews for software engineering students.
A team of researchers from academia has published a paper detailing an innovative educational tool that uses Generative AI to simulate stakeholder interactions for software engineering students. The method, designed by Eva-Maria Schön, Michael Neumann, and Tiago Silva da Silva, addresses a core challenge in agile development: the practical difficulty of involving real users and customers. Their solution employs a structured meta-prompt to create AI Personas that act as simulated stakeholders, allowing students to practice conducting requirements-gathering interviews in a controlled, repeatable environment.
Following the AI-led interviews, students apply established agile practices—such as story mapping and impact mapping—to document the elicited requirements. Crucially, the learning process concludes with a structured group discussion where students reflect on the technical capabilities and ethical limitations of using tools like GPT-4 or Claude 3. The researchers report that this approach, used over multiple academic terms, successfully provides hands-on experience with modern requirements engineering while fostering critical thinking about AI's role. A key technical advantage is the use of a provider-agnostic meta-prompt, which ensures the teaching framework remains flexible and is not locked into a single large language model.
- Uses a meta-prompt to create AI Personas that simulate real project stakeholders for student interviews.
- Students then apply agile practices like story mapping to document requirements based on the AI interactions.
- The method is LLM-agnostic, tested over several terms, and includes reflection on AI's technical and ethical limits.
Why It Matters
This bridges the gap between theory and practice, training the next generation of developers to use and critically evaluate AI tools in real-world workflows.