Active Information Selection: Visual Attention Through the Hands.

Area of Science:

• Cognitive Science
• Developmental Psychology
• Artificial Intelligence

Background:

• Understanding how learning systems process real-world uncertainty is crucial for both human and artificial intelligence.
• Natural intelligence systems, like human toddlers, interact with their environment through perception-action loops.
• A child's actions critically shape their sensory perception, influencing the information they receive.

Purpose of the Study:

• To investigate how perception-action loops in toddlers simplify the learning environment.
• To describe the visual learning environment from a young child's perspective during real-time play.
• To measure visual information perceived by children during interactive toy play.

Main Methods:

• Novel methodology to analyze the visual learning environment from a child's viewpoint.
• Real-time measurement of visual information during toddler-parent toy play.
• Analysis of perception-action loops and their effect on information selection.

Main Results:

• Visual perception is primarily driven by the child's own actions and secondarily by social partners' actions.
• Manual actions are key in creating focused visual experiences, highlighting specific objects.
• Action-driven filtering of visual input provides constrained, clean data, likely facilitating cognitive learning.

Conclusions:

• Perception-action loops actively shape the learning environment for young children.
• Toddlers' actions play a vital role in selecting and filtering sensory information for learning.
• Findings offer insights into natural learning mechanisms applicable to artificial intelligence development.