HiTA: A RAG-Based Educational Platform that Centers Educators in the Instructional Loop

Large Language Models (LLMs) have the potential to influence the current educational framework by providing answers to students without educators’ involvement. Herein, we have developed a new system, called HiTA, that places educators at the center of AI-assisted learning loops. The key concept involves using LLMs as teaching assistants to amplify educators’ expertise with their pedagogical supervision. Built on retrieval-augmented generation, HiTA incorporates advanced features such as a user-friendly interface, a multi-mode prompt processing subsystem, and a customization subsystem that provides course-specific support. Compared to general LLMs, HiTA provides pedagogically oriented and course-adherent responses. Our system has been deployed to 6 educators and 400 students spanning entry-level to advanced courses at the Colorado School of Mines for two semesters. In the Fall 2023 semester, 270 students posted over 14,000 questions within approximately two months. From the 47 feedback forms collected post-semester, over 97% of survey respondents found HiTA helpful, and more than 80% considered it more beneficial than ChatGPT.