A stochastic world model on gravity for stability inference

Area of Science:

• Cognitive Science
• Neuroscience
• Physics

Background:

• Humans intuitively understand object stability and gravity.
• Previous research suggests a 'world model' aids in mental simulations of physical phenomena.
• The precise nature of this internal world model for gravity remains unclear.

Purpose of the Study:

• To investigate the internal world model of gravity in humans.
• To determine how this model represents gravitational forces.
• To link the model's characteristics to subjective perceptions of object stability.

Main Methods:

• Measured participants' sensitivity to gravity.
• Developed a computational model incorporating reinforcement learning.
• Compared model predictions with observed human behavior and subjective reports.

Main Results:

• The human world model of gravity is not a perfect replica of physical laws.
• It encodes gravity's vertical direction as a Gaussian distribution (stochastic representation).
• This stochastic model explains subjective stability judgments and the illusion that taller objects are more likely to fall.

Conclusions:

• The stochastic nature of the gravity world model likely arises from experience-dependent learning.
• This representation balances accuracy and speed for efficient stability inference.
• The findings illustrate how a priori knowledge is implemented in the mind for flexible environmental navigation.