Inferring mass in complex scenes by mental simulation.

Area of Science:

• Cognitive Science
• Physics Perception
• Human-Computer Interaction

Background:

• Humans intuitively understand object properties like mass through interaction.
• Perceiving physical attributes is vital for learning about the world.

Purpose of the Study:

• To investigate people's ability to reason about relative object masses in realistic 3D environments.
• To develop and validate a cognitive model explaining these mass inferences.

Main Methods:

• Conducted three experiments involving naturalistic 3D scenes and object collisions.
• Collected human judgments on relative object masses.
• Developed a computational model integrating Bayesian inference and approximate Newtonian physics.

Main Results:

• Participants accurately inferred relative object masses.
• Inferences were refined and improved over time with continued exposure.
• The proposed cognitive model accurately predicted human judgments.

Conclusions:

• Humans possess a robust capacity for inferring physical properties, specifically mass, from observed interactions.
• A cognitive model combining probabilistic inference with approximate physics simulation can explain these intuitive physical reasoning abilities.