Signal Cite or Decline: A Strict Course-Grounded Chatbot for STEM Lecture Videos

Summary

Recorded lecture videos, often enhanced with search and summarization features, are a standard study resource, but students cannot easily ask course-specific questions or verify answers against an instructor's lecture. The researchers report a semester-long deployment of the VideoPoints platform with a retrieval-augmented chatbot that answers from course lecture materials and returns timestamped citations, retrieving only from the active course and using chapter summaries to guide transcript ranking. Across 833 messages, 70.5% included citations, none crossed a course boundary, and when no lecture evidence matched, the chatbot usually declined rather than answering. Among users, citations were the most consistently useful feature, while practice-question generation was the strongest unmet request. The design was also evaluated on the real-world test split of EduVidQA, a public multimodal benchmark for lecture-video question answering, improving correct-lecture retrieval by 6.3 percentage points over dense-only retrieval, showing that effective deployment depends on course isolation, supported citations, and alignment with students' study practices.