Publication
Reflective writing, a valuable educational practice, helps learners analyze their thoughts and experiences. However, learners often struggle to translate their experiences and thoughts into written form. Recent advancements in large language models (LLMs) have demonstrated their potential for writing assistance and learning support, but they lack personalization, as their responses are not typically grounded in the personal experiences and prior knowledge states of individual learners. Retrieval-augmented generation (RAG) offers the potential to enable LLMs to incorporate learners’ past data to provide personalized support. Despite potential, RAG has yet to be fully evaluated for reflective writing. In this late-breaking work, we introduce Memoire, a reflective writing assistant using RAG to provide learners with personalized suggestions. We evaluate the effectiveness of our tool on 100 students in a classroom study, comparing three types of suggestions identified in prior research. Our study provides early insights into the potential of RAG-based methods to support reflective writing among students.