Designing and Evaluating AI-Based Reflection Support Systems for Students

Metacognition, learners’ ability to regulate their own learning, is considered central to effective learning. However, many struggle to engage in deep and structured reflections. Although AI-based writing assistants and chatbots have been increasingly used to support reflection, existing tools often rely on generic prompts, surface-level feedback, and static interaction modalities, with limited grounding in learning theories or contextualization in each learner’s personal experiences. My research aims to investigate how AI-powered support modalities can be designed to support reflection. Drawing on principles and theories from learning sciences, HCI, and NLP, I design and evaluate AI-based paradigms that provide structured, theory-informed, useful, and personalized reflective support. In my ongoing thesis work, I plan to further investigate and refine AI-powered interaction modalities for reflection, and develop principled ways of integrating prior experiences of learners from different educational levels into reflective support.