The UNCAGE project, led by Ujwal Gadiraju, set out with an ambitious goal: to explore how conversational agents could help students develop better study habits and improve their mental well-being. These are both pressing issues in higher education, especially after the pandemic made digital forms of support more relevant than ever. Before UNCAGE officially began, two tools had already been developed. Goalkeeper helped students set SMART goals for studying. Trainbot supported mentors in practising Motivational Interviewing—a technique often used in mental health care. Both Goalkeeper and Trainbot are rule-based systems. That is: they follow pre-set scripts. Since then, large language models (LLMs), like ChatGPT, have changed the field significantly. UNCAGE work directly with LLMs, but new research lines involving them have grown from the project.
Aiming for impact
The aims of UNCAGE were twofold. First: to personalise conversational agents to support students' learning in a more engaging way. Second: to use these agents to address mental health issues more directly. In a general sense the project laid a foundation, one that others might build on and further implement or adopt within TU Delft or other universities.
Seeds for future work
Concrete outcomes of UNCAGE sparked new research directions. One of these looks at how chatbots can support teamwork, by making group contributions more visible. Another focuses on how the framing of a chatbot affects how users trust and interact with it. There’s also ongoing work into human-centred explainable AI, which deals with how non-expert users often place too much trust in seemingly helpful systems. UNCAGE tackled real issues of student stress, motivation, and the growing role of digital support systems. It also fits within a wider research trend that looks at how AI can support human learning and well-being.