Predictive looking and predictive looking errors in everyday activities

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Modeling

Background:

• Human experience is naturally segmented into discrete events.
• Event comprehension models suggest prediction errors trigger segmentation.
• Direct measurement of prediction errors during ongoing comprehension is lacking.

Purpose of the Study:

• To investigate the role of prediction error in event segmentation.
• To measure online prediction errors using eye-tracking.
• To validate eye-tracking as an assay for prediction error.

Main Methods:

• Participants watched everyday activity movies while their eye movements were tracked.
• Gaze location was used to predict future hand locations.
• Gaze-based prediction error was compared with a computational model's error and uncertainty.

Main Results:

• Gaze-based prediction error correlated better with model error than model uncertainty.
• Eye-tracking successfully measured online prediction error.
• Gaze-based prediction error was significantly correlated with behavioral event segmentation.

Conclusions:

• Predictive looking provides a valid measure of online prediction error.
• Prediction error is closely associated with the process of event segmentation.
• Findings support theoretical models linking prediction error to event comprehension.