Inspiration
We noticed that lecture videos are often passive and inefficient, students lose focus, but neither they nor instructors know exactly when or why. We wanted to make attention measurable and actionable using AI.
What it does
Our project analyzes lecture videos using LLMs and cognitive modeling to detect when attention drops. It highlights confusing segments, pacing issues, and cognitive overload points, giving both students and educators clear insights to improve learning.
How we built it
We used TribeV2 and Mirofish to process lecture video transcripts and metadata. An LLM analyzes semantic complexity, pacing, and topic shifts, while a brain-inspired model estimates attention levels over time. We then visualize engagement dips and key problem areas.
Challenges we ran into
One major challenge was accurately modeling “attention”, since it’s not directly observable. We also had to align LLM outputs with time-based video segments and ensure our signals weren’t just noise but meaningful patterns.
Accomplishments that we're proud of
We successfully built a system that maps attention over time and produces actionable insights. The integration of LLM reasoning with cognitive modeling to simulate engagement is something we’re especially proud of.
What we learned
We learned how to combine AI with human-centered design, especially how difficult it is to quantify cognitive states like attention. We also gained experience working with multimodal data and aligning models with real-world use cases.
What's next for Project Clarity
We want to incorporate real user data, improve accuracy, and build a real-time feedback system for live lectures. Long term, we aim to personalize learning experiences at scale.
Log in or sign up for Devpost to join the conversation.