Children's scale errors are a natural consequence of learning to associate objects with actions: A computational...

Area of Science:

• Cognitive Development
• Computational Neuroscience
• Developmental Psychology

Background:

• Young children exhibit scale errors, inappropriately acting on objects based on size.
• Existing theories attribute scale errors to immature action planning, object representation complexity, or teleofunctional bias.

Purpose of the Study:

• To computationally model children's learning of object-action associations and action selection.
• To investigate the developmental origins of scale errors in children using computational methods.

Main Methods:

• Developed a Developmental Deep Model of Action and Naming (DDMAN) based on dual-route action selection theory.
• Emulated learning processes where the model associates objects with actions, considering both shape and size.
• Analyzed the model's error patterns and the emergence of object-action associations during training.

Main Results:

• The DDMAN model replicated scale errors observed in young children, with error rates decreasing over training but not vanishing.
• The model initially organized object-action associations primarily by shape, leading to size-independent action selection.
• With increased experience, the model learned to integrate object size into its action selection process.

Conclusions:

• Scale errors in children are a natural outcome of the learning process for associating objects with actions.
• Computational modeling provides insights into the developmental trajectory of action selection and error correction.
• The findings suggest that learning to utilize object size alongside shape is a gradual developmental process.