When radiologists compare notes, their eyes teach AI to see better

It's like the difference between one person marking up a chest X-ray with a highlighter versus a panel of experts doing it together — the computer learns from where *everyone* looked, not just one person's gaze.

This means AI trained on collaborative human attention outperforms solo-expert systems by 5–7 percentage points, and you only need three radiologists' eye-tracking data to capture most of the benefit.