AiGet
AI-Assisted Hidden Knowledge Discovery on Smart Glasses|
Highlight
🔍 Have you ever noticed the hidden surprises around you? 📸
—— How Smart Glasses Reshape Your Curiosity ——
Our latest research, AiGet, will be presented at CHI 2025! ✨
👉 Transform AI into an "environment detective" through proactive perception with smart glasses:
- * Real-time analysis of your gaze and surroundings
- * Providing unexpected but insightful knowledge
- * Personalized recommendations relevant to daily life
Interaction Flow
We support a mixed-initiative method to enable both self-directed (user-initiative) and unintentional (AI-initiative) informal learning. The interaction flow consists of 5 stages.
User's View Without Glasses
Mixed-Initiative LLM Trigger
FPV + Time Change
AI-initiated Query - Constant Sensing
Triggered Here
Gaze Fixation
AI-initiated Query - Implicit Gaze
Ring Mouse + Verbal Query
User-initiated Query
Comparison with Baselines
We conducted a ablation study by comparing the AiGet pipeline with two baselines:
(1) Baseline w/o R : a baseline removes the Rules we used, while keeps the User Profile.
(2) Baseline w/o RP: a baseline removes both Rules and the User Profile.
We demonstrate two scenarios below to show the user needs of using rules and user profile to 1) balance both real-time intention and long-term interest, and 2) providing "surprise" at suitable moments.
First-Person View (FPV)
User Interests: Alcohol & Japanese Culture