We are thrilled to share the highlights from a demo session at the International Society of Learning Sciences 2024 (ISLS2024). It was a milestone for our team as we unveiled our Retrieval-Augmented Generation (RAG) - enhanced chatbot, "LS Explorer," powered by the GPT-4o model.
Introducing LS Explorer
LS Explorer is designed to make navigating the extensive body of research published in ISLS conferences since 1995 more accessible and intuitive. Our chatbot offers three distinct functions to cater to diverse user needs:
Synthesis LS Research
Chat with One Paper
Synthesis with Memory
LS Research Synthesis
This function enables users to ask the chatbot any questions related to learning sciences research. Whether you're defining concepts, measuring constructs, or exploring research methodologies, LS Explorer provides comprehensive answers backed by three to five references. Each reference includes real links to the ISLS repository, ensuring users can verify and delve deeper into the sources. This feature is particularly beneficial for new researchers, educators, and ed-tech developers, making LS research more accessible and user-friendly.
Chat with One Paper
Our second function allows users to engage in a detailed conversation with a specific paper. By searching for a paper using keywords, author names, or the full title, users can access and explore the content in-depth. This personalized interaction simplifies the research process, helping users to focus on the material most relevant to their interests.
Synthesis with Memory
An advanced version of the Synthesis LS Research function, Synthesis with Memory equips the chatbot with the ability to remember and contextualize previous interactions. This feature allows users to ask follow-up questions and dive deeper into topics without starting from scratch. It's a powerful tool for anyone looking to build a comprehensive understanding of specific research areas over multiple sessions.
Positive Reception and Feedback
The response to our demo session was positive. Participants praised LS Explorer for its ease of use and accurate referencing. These affirmations highlight LS Explorer's potential to transform the research landscape by making it more approachable and reliable. The reduced occurrence of hallucinations, a common issue with AI models, was particularly appreciated by the attendees.
Constructive Feedback and Future Improvements
In addition to the praise, we received valuable constructive feedback. Participants suggested combining fine-tuning with the RAG approach to enhance the chatbot's performance. Another suggestion was to scale this tool to a wider scientific domain, expanding its utility beyond learning sciences.
Looking Ahead
The feedback and enthusiasm we received at ISLS2024 have energized our team. We are committed to incorporating these insights to refine our RAG approach.
Ready to see how Curiously™ can elevate your classroom? Request a demo today and discover how personalized learning at scale can become a reality for your students!
