Where to look next? Eye movements reduce local uncertainty

Area of Science:

• Cognitive Neuroscience
• Computational Vision
• Psychology

Background:

• Active vision involves dynamic eye movements to acquire task-relevant information.
• Information theory offers a framework to understand how visual information, prior knowledge, and goals guide eye movement planning.
• Understanding visual search strategies is crucial for explaining human perception and interaction with complex environments.

Purpose of the Study:

• To investigate the strategies observers employ when planning sequential eye movements in a controlled visual task.
• To differentiate between global information maximization and local information reduction as potential eye movement planning rules.
• To computationally model and analyze fixation placement to uncover underlying visual search strategies.

Main Methods:

• Measured human eye movements during a shape-learning and -matching task with tightly controlled visual information.
• Employed computational models to analyze sequential fixation data.
• Compared observed fixation patterns against predictions from information-theoretic strategies (global vs. local uncertainty reduction).

Main Results:

• Observer behavior, while seemingly aimed at maximizing total information gain (reducing global uncertainty), was more accurately explained by a local strategy.
• Analysis of sequential fixation placement revealed a tendency to fixate locations that maximally reduce local uncertainty.
• This suggests a preference for immediate information gain at each fixation point.

Conclusions:

• Human eye movement planning in active vision may rely on a local uncertainty reduction strategy rather than a global one.
• This finding provides insights into the cognitive mechanisms underlying efficient visual information gathering.
• The study highlights the importance of analyzing sequential fixation patterns to understand complex perceptual strategies.