Gaze Analysis
Read attention and eye-contact patterns from recorded speaking practice.
The problem
Attention is the hardest part of presence to self-assess. Speakers rarely know where they were actually looking, or for how long.
The idea
Extract gaze and attention signals from a recording and describe them back plainly, without judgment.
What it does
- Estimates gaze direction and eye-contact duration from video.
- Summarizes attention patterns over the course of a response.
- Pairs with interview or presentation practice sessions.
How it works
- 1
Capture
A short recording is made on camera.
- 2
Track
Gaze direction is estimated frame by frame.
- 3
Summarize
Patterns are aggregated into readable signals.
- 4
Compare
Repeat attempts are placed side by side.
Research connection
Classroom Engagement — attention as an observable, improvable signal.
See the research pillarsDesign decisions
- Descriptive language only; no personality inference.
- Signals stay with the student, never the instructor.
Behavioral design
Making an invisible behavior visible is usually enough to change it without any instruction.
Classroom use
Prototype paired with interview practice and presentation rehearsal.
What I learned
- Students trust attention data far more when it is never graded.
- Raw numbers mean little; the comparison between two attempts is what teaches.
Related work