I’m also on the job market! If you’re working on predictive processing, event cognition, or computational modeling — or know of exciting postdoc opportunities —please DM!
📄 Paper: psycnet.apa.org/record/2026-...
🙏 With: Matt Bezdek, Tan Nguyen, Chris Hall and @jzacks.bsky.social

Next, we’re digging deeper into richness of free-viewing gaze patterns — linking embeddings from large visual models to eye movements to see how meaning guides where people look.
More soon! 👀

Gaze predicted the actor’s hand positions up to 9 seconds ahead. Gaze errors matched a computational model’s prediction errors, and when predictions errors increase, people were more likely to mark event boundaries.

To test these predictions, we tracked where people looked while they watched everyday activities.
Using a model comparison approach and multivariate regression models, we estimated how far ahead their eyes predicted the actor’s hand movements—and quantified the magnitude of those prediction errors.

Event Segmentation Theory suggests that we’re constantly making predictions, and that prediction errors serve as key signals for event boundaries.
However, it’s challenging to directly quantify people’s predictions and prediction errors during natural, everyday activities.