Mind Games: Game Engines as an Architecture for Intuitive Physics.

Area of Science:

• Cognitive Science
• Developmental Psychology
• Artificial Intelligence

Background:

• Human intuitive physics understanding is complex and not fully explained by existing models.
• Young infants demonstrate surprising physical knowledge, posing a challenge for developmental theories.
• Artificial intelligence (AI) systems often lack human-like common sense reasoning abilities.

Purpose of the Study:

• To propose and explore the mental physics engine hypothesis for intuitive physical inference.
• To unify phenomena in infant intuitive physics and explain 'physics illusions'.
• To provide a basis for computational models of infant physical knowledge development and inform AI common sense.

Main Methods:

• Analogy drawn between machine physics engines in video games and human mental representations.
• Focus on intuitive physics in young infants as a key test case.
• Explanation of 'physics illusions' and implications for AI development.

Main Results:

• The mental physics engine hypothesis offers a unifying framework for infant physical reasoning.
• The hypothesis explains previously puzzling infant behaviors and 'physics illusions'.
• Provides a potential computational basis for understanding the development of physical knowledge.

Conclusions:

• Human intuitive physics may operate via a mental physics engine analogous to game engines.
• This framework advances our understanding of infant cognitive development and physical knowledge acquisition.
• Offers insights for creating AI with more robust common sense reasoning capabilities.