A cognitive category-learning model of rule abstraction, attention learning, and contextual modulation

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Category learning is fundamental to cognition.
• Existing models struggle to explain rule acquisition from scratch and individual differences.

Purpose of the Study:

• Introduce the Category Abstraction Learning (CAL) model, a cognitive framework for category learning.
• Explain how rules are learned from scratch based on novel assumptions.
• Capture systematic and individual differences in learning paradigms.

Main Methods:

• Formalized a cognitive framework integrating similarity-based generalization and dissimilarity-based abstraction.
• Incorporated two attention learning mechanisms and error-driven knowledge structuring.
• Simulated benchmark tasks like the Six Problems and 5-4 problem.

Main Results:

• The CAL model explains rule emergence from stimulus generalization and category contrast.
• Attention mechanisms focus learning on rules or error-producing contexts.
• Model accounts for partial rule application and individual differences in rule extrapolation.

Conclusions:

• The CAL model offers a novel account of category rule acquisition from scratch.
• It explains diverse phenomena challenging existing learning theories.
• CAL provides a framework for measuring cognitive processes like attention and abstraction.