Designing Ethical AI Companions for Software Engineering Education: Lessons from a Supervised Chatbot for UML Learning

This chapter presents the design and deployment of an ethical, locally executed chatbot to support the teaching of conceptual modeling with Unified Modeling Language (UML) in undergraduate software engineering courses. Developed and implemented at the University of Salamanca during the 2024–2025 academic year, the system integrates a generative model (DeepSeek-v2) through a Django-based platform running entirely on institutional servers. Its distinctive architecture combines privacy-by-design principles with teacher-supervised interactions, ensuring that all exchanges remain anonymized while fostering a culture of critical reflection on the responsible use of artificial intelligence in higher education. The pilot involved 13 students who collectively generated 110 interactions with the system, providing qualitative evidence of the tool’s potential for autonomous learning, usability, and ethical awareness. Quantitative usability scores (mean 3.75/5) and user feedback indicate promising acceptance, though challenges were observed in maintaining adherence to AI documentation requirements and in managing computational load under concurrent use. Beyond its technical contribution, the chapter discusses the broader pedagogical implications of integrating locally hosted AI assistants in software engineering education. It argues that ethical and sustainable adoption of AI requires not only robust infrastructures but also explicit curricular frameworks that promote digital responsibility and reflective practice among students and instructors alike.