The Subconscious Director: Dynamically Personalizing Videos Using Gaze Data
summary
A webcam-based system called eyeDirect records gaze and computes nine fixation-based features per area of interest to infer which on-screen element a viewer prefers, then selects the next scene. A between-subjects study with N = 175 compares control, random branching, and gaze-driven continuations using naive majority voting versus machine learning classifiers. Naive prediction fails to boost engagement, while ML improves prediction accuracy and positive personalization raises focused attention and involvement but not novelty.
AI-written summary · about
related talks
17mGaze Augmentation in Egocentric Video Improves Awareness of Intention
17mLearning Cooperative Personalized Policies from Gaze Data
23mParallel Eyes: Exploring Human Capability and Behaviors with Paralleled First Person View Sharing
14mTransform intent into action through your eyes
17mMise-Unseen: Using Eye Tracking to Hide Virtual Reality Scene Changes in Plain Sight
19mDynamic Stereoscopic 3D Parameter Adjustment for Enhanced Depth Discrimination
43mThe Evolution of User Interfaces
34mHow Eyetracking can be Beginning of the End of Privacy