Robustness of gaze signals for cognitive-state modeling under AI advice conditions

Determine whether eye-tracking signals, including gaze behavior and pupil measures, provide a robust basis for modeling users’ cognitive states when the decision-making context is influenced by AI advice, including the presence of assistance and the correctness of the advice.

Background

Eye tracking is widely used as a non-invasive source of behavioral correlates for cognitive and attentional effort. However, in AI-assisted decision-making, the context can shift substantially depending on whether AI advice is shown and whether that advice is correct, which may alter how gaze and pupil signals relate to underlying cognitive states.

The paper motivates the need to test whether these physiological and behavioral signals remain reliable indicators for cognitive-state inference when AI assistance is introduced, potentially changing users’ attention allocation and strategies.