Attention Check

A method and a small viewer for checking AI-predicted attention maps against real eye-tracking data on the same screens.

The setup

Twenty to thirty interface screens, desktop and mobile. The model predicts where people will look; the lab records where they actually do. Both maps are compared on the same areas of interest.

What to expect

Agreement on faces, large type and strong contrast. Disagreement where meaning matters more than salience: the small link someone is actually searching for.

Status

Offered as an RTU bachelor thesis topic for 2026/27.